Comparing Drivers of GHG Emissions at Different Industrialization Stages: Structural Decomposition Analysis for US, China and India

Comparing Drivers of GHG Emissions at

Different Industrialization Stages:

Structural Decomposition Analysis for US,

China and India

Han Lin (h.lin.3@student.rug.nl)

Student No. S2883082

Faculty of Economics and Business, University of Groningen

Supervisor: prof. dr. mr. C.J. Jepma

[Abstract] As environmental problems are being increasingly prominent, greenhouse gas

emissions (GHG) are drawing more and more attention. This paper aims to compare the drivers of GHG emissions at different industrialization stages during 1995 and 2009, taking US, China and India as examples for typical industrialization levels. Based on analysis for world input-output table and structural decomposition analysis (SDA), results prove that the levels of industrialization make drivers of GHG emissions different.

[Key Words]Industrialization; Structural Decomposition Analysis; Input-Output; GHG Emissions

1. Introduction

There is no doubt that a series of natural phenomena resulted from global warming, such as melting glaciers and rising sea levels, are threatening the whole ecosystem and even the surviving of our human beings. In essence, the most effective way to slow down global warming is reducing the greenhouse gas (GHG)1 _emissions.

However, GHG emissions are never changed for no reason during the development of economies. Especially, the occurrence of industrialization is always seen as a turning point in the social development and the social changes happened at that time are closely relate to the evolution of GHG emissions. Since the development of industrialization also experiences different stages, in this paper, the relationship between industrialization in three countries and GHG emissions is to be revealed by applying Structural Decomposition Analysis (SDA) to compare contributions of factors that changed in different industrialization stages to GHG emissions during 1995 and 2009.

It is also reasonable to choose US, China and India as objects in this study since not only they are the largest GHG producers in the world, but also they can best illustrate the relationship between industrialization and emissions. More specifically, when the whole world is urging for reducing emissions, India’s environment minister stated in 2014 that the eradication of poverty was of top priority for India and the country would not cut GHG emissions since they tend to industrialize and grow faster. While for China, at the United Nations Climate Change Conference in 2015, even though 1 _{According to the United Nations Framework Convention on Climate Change (UNFCCC), greenhouse gas}

indicates the direct greenhouse gases such as CO2- Carbon dioxide, CH4- Methane, N2O Nitrous oxide, PFCs

-Perfluorocarbons, HFCs - Hydrofluorocarbons, SF6- Sulphur hexafluoride, also indirect greenhouse gases such as

we promised to limit emissions, it is unavoidable to peak emissions by around 2030 and it will be an

important period for the completion of China’s industrialization2_{. Besides, the United States as a}

country which had fully industrialized before the 21st_{century is also to blame since it emitted most}

GHG in the history. Therefore, it is meaningful to analysis the emissions of these countries from the perspective of their threat to the earth.

Besides, levels of industrialization of three countries are extremely typical, as US had already industrialized and is going through de-industrialization, China is still industrializing while India seems haven’t started yet. Since three countries were going through completely different stages of industrialization during the same years (1995-2009), it would make it possible to find out the evolution of emissions along the industrialization process by comparing the decomposition results for the three countries.

The remainder of this paper is organized as follows: section 2 reviews plenty of previous relevant studies, followed by detailed description of industrialization stages that US, China and India were going during the research years in section 3. After that, main methodology will be introduced in section 4, together with the data processing method in section 5. Then, results from structural decomposition analysis will be carried out in section 6, taking industrialization related issues into account. Finally, conclusions will be covered.

2. Literature Overview

Since environmental problem has widely attracted public attention, massive research focusing on such issue has been carried out by scholars decades ago. There is a prominent concept called the Environment Kuznets Curve (EKC) which was firstly introduced in the work of Grossman and Krueger (1991) which focused on the environmental impacts of NAFTA and was based on the background study for 1992 World Bank Report, done by Shafik and Bandyopadhyay (1992).

The EKC intuitively depicts the relationship between emissions and the development of economies. More precisely, as is shown in figure 1, the environment gets polluted as economic growth

2 _{An Exploration into China s Economic Development and Electricity Demand by the Year 2050 , by Zhaoguang}

initiates and there will be a turning point where the environment starts to recover as the economy further develops.

There were also scholars give different explanations to such phenomenon. Proponents of the EKC hypothesis argue that ‘at higher levels of development, structural change towards information-intensive industries and services, coupled with increased environmental awareness, enforcement of environmental regulations, better technology and higher environmental expenditures, result in leveling off and gradual decline of environmental degradation3_{.’ Also, as Stern (2003) further broadened the}

concept of EKC to world level, he explained that ‘In rapidly growing middle income countries the scale effect, which increases pollution and other degradation, overwhelms the time effect. In wealthy countries, growth is slower, and pollution reduction efforts can overcome the scale effect’. What’s more, as he comprehensively summarized EKC related studies, he pointed out that the EKC can be explained by several ‘proximate factors’ such as changes of output, input and technology which involves both production and emission efficiency.

A sizable amount of research on relationship between environment and development has been done based on EKC. Brodie (1905) carried out an early work concerning that the frequency of foggy days in London in the late 19th_{century was closely related to its industrialization. He collected weather}

related data between 1871 and 1903, then showed them as an inverted U-shape graph where the annual number of foggy days in London firstly rose since 1870s and then fell steadily after 1890s. As Brodie believes, such phenomenon has direct connection with coal burning: a huge amount of energy was needed when industrialization was realized, while the weather in London turned better only when further technological, legal and social changes were made. However, many scholars questioned the official definition of ‘foggy day’ as well as the reliability of the data in Brodie’s work. Clay and Troesken (2010) then applied larger data and more other indicators to reconsider Brodie’s interpretation. As they found, it was not only the technological change which mentioned by Brodie but also the redistribution of population, successful policies and the use of new energy (gas) that contributed to the reduce of the number of foggy days in London. What’s more, they proved that such an inverted U-shaped graph in Brodie’s work was exactly the prototype of EKC even though nobody named it at that time.

Besides, A lot of effort was also made to reassess the EKC. Bradford et al (2000) reanalyzed previous studies on EKC. Their work was based on Grossman and Krueger(1994)’s research but applied the new specification to augment their data description. Taking the econometric problem which stems from using cross-sectional data into consideration, authors used fixed effects models to make their conclusions more plausible and they found some support for EKC for some pollutants while some rejections in other cases.

Similar conclusions were also carried out by Georgiev and Mihaylov (2015), who also dedicated to reassessing the existence of EKC by applying panel data for 30 OECD countries, and the main air pollution that they concentrated on consists of four local (SOx, NOx, CO, VOC) and two global (CO2,

GHG) air pollutants. As they individually made researches for all these gases and graphed their relationship with income growth, they found that the inverted U-shaped curve does not hold for all gases. Besides, LEE et al (2009) reexamined EKC for CO2emissions only, but in different countries.

Their conclusion rejected ‘one fits all’ hypothesis and they found such EKC holds in certain income level and regions but not in others.

In fact, the concept of EKC is extended from Kuznets Curve (Kuznets,1955) which is an inverted U-shaped curve explaining that the inequality in an economy firstly increases as the economy develops and starts to decrease at certain point of development. The reason why I refer to this is that the mentioned ‘development’ in Kuznets Curve is originally related to industrialization, in other words, the development level is reflected by the degree of industrialization in Kuznets’s work. Therefore, there is probability that all the works above regarding emissions and development should also be more or less applicable to researching the relationship between emissions and industrialization.

prominent finding of the paper is the occurrence of the N-shaped income-pollution relationship (figure 2). According to their analysis, the falling arm of original EKC is not that realistic in the real world since it happens only when the manufacturing share keeps decreasing infinitely but never reach zero. Alternatively, a new rising arm seems more reasonable. Therefore, based on such finding, authors give new interpretation to N-shaped curve: the original EKC which indicates that the environmental problem will be automatically solved as the income gets high enough is misleading and unsustainable, there will be a turning point where the income-pollution relationship turns positive again.

Further more, there are also many studies focusing on quantitative relationship between emissions and industrialization. Cherniwchan J (2012) developed a model based on simple neoclassical model and small open economic environment and used rich data for 68 countries over the period 1970-2000 to examine how industrialization affects the environment. Apart from that, Pan J (2004) carried out a study regarding the industrialization, energy use and GHG emissions in China, mainly indicated that energy supply can constraint industrialization, and as a result, the mitigation of emissions can affect the industrialization level. Besides, Schipper et al (1997) investigated the relationship between CO2

emissions and pure energy use. They observed that during 1970s and 1980s in 10 industrialized countries, two main factors, namely decreased end-use energy intensities and carbon content of energy, contributed to the declined energy use related CO2 emissions. However, by the early 1990s, the

situation was reversed as GDP and energy services activity continued to grow. Such finding draws particular attention to policy changes and technological innovation through which the energy use related CO2emissions can be mitigated.

It is undoubtedly that much more relevant studies can be found. So far, however, there are some concerns worth thinking.

Firstly, there is little work focusing on the emissions throughout the complete stages of industrialization, for example, namely pre-industrialization, industrializing and de-industrialization. Some may have tried to do so for a specific country, however unfortunately, the whole process of industrialization takes so long that the required data can be very limited.

decomposition analysis is needed to track the evolution of emissions. This paper will try to address these issues by using Input-Output Table for 1995-2009. Since there is no database which covers the whole industrialization process of any specific country, alternatively, I will focus on studying emissions in US, China and India individually. But before turning to detailed methodology, there is necessity to distinguish the degree of industrialization in targeted countries.

3. Industrialization in US, China and India

A general phenomenon for industrialization is shown as the rising industrial output proportion and employment rate, therefore in the meanwhile, agricultural output as a proportion of total output keeps decreasing, so does the agricultural employment rate. In this section, introductions for industrialization in three countries will be presented since their differences in industrialization process are main concerns in this research. To briefly but clearly distinguish de-industrialization in US, on-going industrialization in China and premature de-industrialization in India, some individual characteristics will be presented, combined with relevant data.

3.1 The United States

It has been widely admitted that US is one of the most powerful and developed countries in the world. We know that this will never happen for no reason, it was the industrialization in early time that laid the foundation for today’s America. The first industrial revolution in US happened at the end of 18th_{century, just followed British industrial revolution, manufacturing technologies were brought from}

Britain to the United States. In 19th_{century, further industrialization took place with more technological}

innovations such as telegraph and sewing machine, especially following the Civil War, industrialization process began to accelerate and finally accomplished in a century. After that, US approached the stage of post-industrialization, or more precisely, it was and is still going through so-called de-industrialization.

(3) shows the breakdown of GDP 4_{composition in US and reflects an obvious trend of structural}

change. It is clear that since 1980, manufacturing output as a share of GDP has declined continuously while output from service sectors has increased steadily.

Besides, Lawrence and Edwards (2013) carried out a study analyzing American de-industrialization from the perspective of employment. Figure (4) depicts an overall downward trend of US manufacturing employment during 1990 and 2011. To be consistent with the research years (1995-2009) of this work, according to authors’ explanations, international factors which contributed to manufacturing employment decline in US can be plausible since increasing imports from emerging countries like China displaced domestic manufacturing, but more importantly, authors showed more concern to other two main factors, namely productivity growth and demand for goods. They stressed that with the completion of industrialization, improved technologies are able to increase productivity hence further improve welfare level. In the meanwhile, public demand for goods is no longer limited to manufacturing goods, but much more for service goods.

3.2 China

To some extent, it seems unbelievable that China as one of the most ancient countries in the world, its thousands of years history didn’t take itself to the top position in the world, even though there were many famous inventions in ancient time.

Due to relatively closed economy, Chinese industrialization started late in 1953 when the first

five-year plan was implemented. Unlike other countries’ industrialization that initiated from light

industry, Chinese government took an irregular path to firstly develop heavy industry, leading to the consequence that annual growth rate of heavy industry was much higher than that of light industry. One could believe that such ‘advanced’ strategy in China was triggered by its ambitious surpassing consciousness, and the heavy industry did get developed as expected.

Years later, however, drawbacks of the strategy turned out to be obvious when structural problems appeared. Imbalanced relationship between light and heavy industry, accumulation and consumption,

4 _{Agriculture corresponds to International Standard Industrial Classification (ISIC) A-B and includes forestry,}

hunting, and fishing, as well as cultivation of crops and livestock production.

Industry corresponds to ISIC divisions C-F and includes manufacturing (ISIC divisions D). It comprises value added in mining, manufacturing, construction, electricity, water, and gas.

as well as inefficient resource allocation contributed to the fluctuation of economic growth. Industrialization was trapped in a cyclic process as figure (5) shows and statistically proved by unstable manufacturing share in figure (6), thus leading to a lack of sustainable growth ability in national economy. Only by 1990s, China managed to get out of this cyclic process and heavy industry was once again given the leading position. However, different from previous period, returning to heavy industry this time followed the law of normal structural change, and since then, 1990s was the new start for China. Even though we do not observe significant increase of manufacturing share during 1995-2009 since China played as a role of ‘world factory’ and there was little value added in manufacturing sectors, China was still industrializing as the share of agriculture decreased sharply.

3.3 India

As being one of the most ancient countries like China, however, India was not as lucky as China to implement industrialization, situations in India were even worse. Being colonized by the Great Britain for a very long period, India became a complex society and had no opportunity to initiate its own path of industrialization in the early time. After liberalization, the government tried to develop its own industry consciously, but it was not as successful as expected due to overlooking agriculture and light industries, as well as frequent conflicts or wars around its territory. In 1991, revolution happened in India as the new government opened the economy to the world, as we can observe in figure (7), Gross Domestic Product increased dramatically since the beginning of this century. Unfortunately, industrialization was not the reason behind.

Comparing the industrial structure in India, as shown in figure (8), gives answer to the problem. It can be seen that manufacturing has never been the leading industry along with the development of the economy even though India implemented industrialization strategy since liberalization. Alternatively, Indian service industry always takes larger proportion of GDP than that of manufacturing, which indicates that the structural change in India failed to follow general path.

agriculture to other two industries and it is faster for the movement towards service sector, making its employment share higher than that of manufacturing. Besides, we can also observe that by 2010, employment share in agriculture still stayed around 50%, which indicates great importance of agriculture in Indian society. Such phenomena proved failure of industrialization in India, what’s more, those characteristics are exactly in accordance with the performance of so called ‘premature de-industrialization’, which was studied by Rodrik (2015) for low- and middle-income countries, stating that those countries like India ‘are turning into service economies without having gone through

a proper experience of industrialization’.

To summarize, in this section, I briefly introduced three different industrialization stages in US, China, and India, proved by both theoretical and statistical evidence as I firmly believe that development level can be reflected by degree of industrialization, thus making three countries different in several aspects which will influence emissions directly and indirectly.

4. Methodology

4.1 Environmental input-output analysis

The empirical method of this work will be based on input-output analysis following what the Leontief did in his work in 1970 and the original equation is given as:

x=Ax+f (1)

which can be rewritten as

x=(I-A)-1_{f (2)}

where I is an identity matrix, A is a matrix reflects the input relationship among all sectors of the economy. Generally, the item (I-A)-1_{is always replaced by M for simplicity. Another parameter f is the}

When taking GHG emissions into account, the matrix of emission coefficient k 5_{is derived as}

k=w×x-1 ₍₃₎

where matrix w denotes emissions in every sector. Then the total GHG emissions (EMS) to meet the need of final demad f can be computed as

EMS=k×M×f (4)

4.2 Structural Decomposition Analysis (SDA)

As initially carried out by Hoekstra and van der Bergh (2002), SDA has become a standard method for understanding the effect of factors contributing to the change of emissions. In general, a country’s emissions change over time can be commonly decomposed into three factors as the following equation shows:

(5) where EMSstands for the change of emissions,



k

indicates emission efficiency change, M for production structural change (“Leontief effects”, M=(I-A)-1_{),f for final demand and HHdir for}

emissions by households. However, such standard decomposition is far from enough to explore more specific causation. Therefore in this paper, further extensions will be applied, see the equation below:

(6)

In equation (6), as k and M remain unchanged, there are several new parameters that worth noting. Term ysyvpis further extended from f in equation (3). Respectively, indicates the composition of final demand, derived as final demand in each sector divided by total final demand in all countries.

v

y

is the final demand per capita which is calculated as final demand for each sector in every country divided by the country’s population. Finally, p stands for population. Extended decomposition as such will give better explanations for the change of emissions over different industrialization stages.

5 _{Changes in emission coefficient are widely explained by technological changes. See Economic Growth and the}

Environment: An Empirical Analysis (Sander M. de Bruyn)

5. Data and Processing

5.1 Description of Data

In this article, in order to guarantee the consistency of the data source, all the data used is from World Input-Output Database, published by Timmer et al (2015), covering in total 1435 sectors (35 sectors in every country) in 40 countries plus 1 entity for the rest of the world (ROW). In this work, I only focus on the period 1995-2009 since there is limited data in previous years prices (I will come to this later). Such database has a great amount of tables but among which, three types of tables will be applied in my work, namely World Input-Output Tables (WIOTs) in current prices and WIOTs in previous years prices, as well as Emissions to Air Tables. When applying all the relevant data to equations in previous section, it is of great importance to understand how many columns and rows for each matrix, see table (1).

Besides, one more thing that worth noting is the order of operations when calculating equations since matrix operation is entirely different from the operation of numbers. In equation (6), for every part at the right of the equals sign except HHdir (regardless the sign of delta), the operation of parameters should be as follow if parenthesis are used for different operation priority:

(k × (M × (ys× (p × yv))))

5.2 Data Processing

Since decomposition study requires economic data in constant prices thus eliminating the price effects for certain years, data processing is necessary for this purpose. As I specified before, two types of WIOT will be used in this work, the one is in current years’ prices while the other is in previous years’ prices, it is because the two should be combined for the sake of eliminating the price effects.

Choosing 1995 as the base year, every term involves WIOTs and reflects change in equation (6) is not simply the term difference between 1995 and 2009. In other words, for example, we can not derive

k



directly by calculating 1995 1995 2009 2009

x

w

x

w

k







(7)6

where output data in matrix x is all derived from WIOD in current years’ prices.

Alternatively, both WIOTs in current prices and in previous year’s prices should be used simultaneously to calculate the differences for every one year. For example, suppose that I am interested in calculating yearly emission change in the year of 2008 without price effect, then the emission change will be computed as

2008 2009

2008

k

k

k



p





(8)

where

k

2009p indicates 2009’s emission coefficient in 2008’s prices and

k

2008 is in current prices.

Both items are calculated by equation (3) with the same w since emissions are not in price term, but what makes them different is output x: output x in first item is extracted from 2009 WIOD in prices of the year 2008 while x in the other item is derived from 2008 WIOD in current prices. Similarily,

2007

k



is derived by the difference of

k

2008p using 2008 WIOD in prices of year 2007 and

k

2007

using 2007 WIOD in current prices. Such alternative way is applied to all parameters that are calculated by the data from WIOTs, namely



k

,



M

,



y

s and



y

v.

Keeping this in mind, I will further take the average of each term in two polar decomposition as Dietzenbacher and Los (1998) did since they believe that taking the average over two ‘polar’ equations yields results that are generally close to the average over the full set. For example, two polar for emission change in 2008 will be given as follows, and finally aggregate corresponding averaged terms to derive the changes of emissions during 1995 and 2009.

6 _{w is the matrix of GHG emissions.}

polar1: 2008 2009 2008 2009 2008 2009 2008 2009 2008 2009 2009 2009 2009 2009

))))

(

(

(

(

)))

)

((

(

(

)))

(

)

((

(

))))

(

(

)

((

(

)))

(

(

(

)

(

2008 2008 2008 2008 2009 2008 2008 2009 2008 2008 2009 2008 2009 2008 2009 2009 2008 2009 2008 



































































dir v v s v s v s s v s p v s p p

HH

y

y

p

y

M

k

y

p

p

y

M

k

y

p

y

y

M

k

y

p

y

M

M

k

y

p

y

M

k

k

EMS

p p p p p p p p polar2: 2008 2009 2008 2009 2009 2008 2009 2008 2008 2009 2008 2008 2008 2008

))))

(

(

(

(

)))

)

((

(

(

)))

(

)

((

(

))))

(

(

)

((

(

)))

(

(

(

)

(

2009 2009 2009 2008 2009 2009 2009 2008 2009 2009 2008 2008 2009 2009 2008 2008 2008 2009 2008 



































































dir v v s p p v s p p v s s p p v s p p v s p

HH

y

y

p

y

M

k

y

p

p

y

M

k

y

p

y

y

M

k

y

p

y

M

M

k

y

p

y

M

k

k

EMS

p p p p

6. Decomposition Results

6.1 Decomposition Results in Countries

Decomposition results for US, China and India are presented in figures below. However, what draws for particular attention is that the contributions of all factors to GHG emissions in certain countries are calculated at global level since WIOTs consist of the data from all countries in the world, which means not only national changes of their own but also all economic activities over the world account for the variation of GHG emissions in certain countries. Therefore, careful treatments are required for the interpretations of following results.

eliminating GHG emissions in US. Structural change that happened over the world helped to reduce GHG emissions in US by approximately 17% while global changes in technology and consumption patterns had less impacts on GHG elimination (6.8% and 8.8% respectively). Differently, from 2007 to 2009, a rapid drop in GHG emissions can be observed, with the population related contribution decreased by about 4.2% while technological change and structural change in the world further led to elimination of GHG emissions.

Things are different for China and India (figure 11 and 12). In both countries, we clearly see a boost of GHG emissions between 1997 and 2009, with more than doubled (114.4%) in China and nearly doubled (95.8%) in India. For China, there were five factors that simultaneously contributed to the surge of GHG emissions within 15 years, namely global changes of population, industrial structure, consumption patterns, consumption volume and households emissions. Among which, the change of population still to the largest extent affected emissions (118.8%) in China as it did in US, followed by structural change which led to 59.8% increase in GHG emissions. Besides, other three factors made little contributions to increasing Chinese emissions. As the only factor that eased the burden of GHG emissions in China, global technological change steadily led to continuous impact on mitigation of emissions: up to 2009, it had contributed 104.5% reduction to GHG emissions.

While in India, its overall trend was quite similar to that of China, with about 80% increase of emissions resulted from the world population change and no more than 45% increase as a total of changes in consumption volume and patterns, together with households change (25.3%, 11.4% and 6.3% respectively). Global structural change and technological change are two factors that offset a part of GHG emissions in India, namely decreased 4.5% and 22.5% of total emissions.

To repeat, the results presented above are all related to global change of each factor in these three countries but do not help to distinguish specific characteristics of three countries which were experiencing different industrialization stages during the period of 1995-2009. To further prove that it

Figure 11: Contributions of Different Factors to GHG Emissions in China (in ratio), 1995-2009

was the industrialization that made GHG emissions in US, China and India different, following part of discussions will separately compare the contributions of each factor to GHG emissions in three countries so that we can make sure for each comparison, three countries were facing the same level of global changes and the only differences came from their own.

6.2 Comparing Factors Across Countries

6.2.1 Structural Change

The first factor that I would like to compare is structural change (M), for it is the most direct effect of industrialization.

Figure 13 gives a picture of impacts of structural change on GHG emissions in three countries. From which one can observe similar decreasing trend of the contribution in US and India and as I mentioned before, the distinction between two countries embodies their different path of national structural change, that is to say, even though the trends are similar, there are different stories to tell.

In section 3, as I explained, US had already started being de-industrialized since 1995 while India has not even initiated industrialization, instead, it was directly experiencing de-industrialization. However, such de-industrialization was not the same as what happened in US-it was premature. To have a better understanding of how do the two types of de-industrialization make different

contributions to GHG emissions, figure 14 and 15 provide more information. The two figures reflect effects of structural change in US and India at sectoral level, numbers under horizontal axis match with 35 different sectors which are listed in input-output table7_.

In US, there were merely two sectors, namely financial intermediation (28) and other business activities (30), showed little increases in GHG emissions over 15 years while emissions in one half of the remaining sectors stayed unchanged and the other half accounted for approximately 0.7% decreases in total emissions on average (exclude sector 17: electricity, gas and water supply, which will be later compared with that of China). Particularly, there is a significant cluster of mitigation of emissions on several manufacturing sectors (7-12), such phenomenon to a very large extent conform to the structural movement in US: shifting away from manufacturing to services8_.

In India, however, a more complicated distribution of contributions from structural change at sectoral level explains why the de-industrialization happened here is premature. It is not surprising that GHG emissions in agriculture and mining sectors (1 and 2) decreased as the Indian government decided to cut down weights of the two sectors thus preparing for industrialization. What’s more, it is also understandable for increased contributions in service sectors (22-30) and especially in sectors of hotels and restaurants (22) as well as inland transport (23) since India has become a country that heavily rely on tourism. However, different stories happened to manufacturing sectors. In India, structural movement in a large proportion of manufacturing sectors, namely 4, 6, 7, 11, 12, induced the decline of GHG emissions, while only fuel and chemicals related industries (8 and 9), the two so called heavy industries contributed to more GHG emissions. Such diverse contributions reveal the fact that with the declining share of agriculture, service sectors in India firstly get well developed without experiencing a boom in manufacturing sectors, even though the government dedicated to developing heavy industries, such policy was however one-sided to the economy. Further more, decreased emissions from electricity, gas and water supply sector (17) also suggest that being failed to initiate normal industrialization requires less energy for national use.

Differently, however, thanks to the tenth five plan published in 2000, pace of industrialization in China was further accelerated as can be proved by huge increase in GHG emissions due to structural change (figure 13). After that, China became a typical illustration for the concept of ‘industrializing’. A 7 _{For names of all 35 sectors, see table 2 (appendix).}

very prominent characteristic shown in figure 16 is that contributions of structural change to GHG emissions in almost all manufacturing sectors in China more or less increased, even in sector of electricity, gas and water supply (17). This phenomenon reflects the prosperity of manufacturing in China during the period of 1995-2009, and massive energy was required to meet the need of production in manufacturing sectors.

6.2.2 Technological Change

Another factor that gives evidence to ongoing industrialization in China and also largely distinguishes it from (premature) de-industrialization in (India) US is the GHG emissions change caused by technological change. In figure 17, variations of contributions of technological change share similar trend in US and India from 1995 to 2009, with approximately 7% decrease of emissions in US and 22% decline in India. While in China, such trend dropped by more than 100% in 15 years.

Recall the definition of emission coefficient (k) in section 4, where each coefficient is derived by the doing division of total emissions and output in every sector, it measures how much GHG emissions is embodied in one unit of production. It is reasonable that emission coefficient change is widely explained by technological improvement in many previous works as with many advanced and mature technologies are used, productivity in industries will be further enhanced, which means fewer inputs such as capital and especially energy are required to produce the same amount of products as before, therefore less green house gas is emitted given the same quantities of products to be produced.

To further give evidence to different technological changes in US, China and India, a concept named total factor productivity (TFP) should be introduced. To be brief, TFP is always taken as a measure of long-term technological change in an economy. However, it can not be directly measured but is a residual when decomposing the drivers of economic growth in an economy. I’m not going to do more research about TFP, but several relevant statistics are collected from other scholars’ work.

Ozyurt (2009) dedicated to studying TFP growth in Chinese industry during 1952-2005. According to his finding, TFP in China increased by 3.18% from 1993 to 2005. Cardarelli and Lusinyan (2015) found that since 1995, TFP in US firstly went up by nearly 2% in 2000, decreased after 2001 and then increased again until 2004. By 2009, it finally backed to the same level as it was in 1995. Coincidentally, such undulatory pattern perfectly fits the emission trend depicted in figure 17 (blue line). For TFP in India, Das et al (2010) proved that it grew by 1.5% between 1990 and 2004. Even though the research years are not the same in studies above, they will not change the main results. In comparison, TFP in China undoubtedly improved most and its growth was nearly twice as large as that of both US and India, thus no wonder we observe a dramatic decline in GHG emissions resulted from technological change.

Besides, it also draws for attention that there is a difference behind similar trends in US and India. As US had already industrialized, technologies were so mature that the improvement of which met the plateau and it was difficult to make a breakthrough progress. Apparently, however, India had no such advanced technologies. Due to its low starting point, one would expect that even a very small step in technological improvement will have great impacts on GHG emissions mitigation. Unfortunately, there was no enough technological change happened in India during 1995-2009, it was also the performance of premature de-industrialization in India.

6.2.3 Population Change

To further analyze these trends, it is necessary to know how did the population change in three countries. Figure 19 gives a picture of their population growth rates during the research period. As we can see, population in India increased most comparing to that of other two countries, followed by less (around 15%) growth in American population, and the least population growth rate happened to China9_.

Such population growth patterns throw out a question as it is seemingly conflicts with the contribution trends in figure 18: how did the most increase in contribution of population change to GHG emissions happen since China experienced the slowest population growth among three countries?

There are many chain reactions resulted from population growth that relate to GHG emissions, such as deforestation and land/soil degradation (Negdeve, 2002). Besides, a very important reason that helps to explain the paradox above is the change of employment share in sectors. Figures 20 and 21 show changes of manufacturing employment share in China and US individually, figure 9 in this paper shows the situation in India. Figures show that the share of manufacturing employment in China reached more than 40% in 2009 comparing to 20% in 1995, while such share in India increased merely 5% in 15 years. Differently, the share steadily decreased in US. The appearance of these phenomena is closely related to their different industrialization stages.

Even though China had the lowest rate of population growth, the absolute number of population growth was still large. As China is experiencing industrialization, more labor force provided by lager population is able to meet the demand by booming manufacturing industry to produce more products. In the meanwhile, energy as an essential part of production is also increasingly needed, hence GHG 9 _{Mainly due to the one-child policy, population growth in China was largely limited.}

emissions will inevitably surge. Similarly, as India was going through premature de-industrialization, there was little development in its manufacturing sectors, therefore only a small part of manufacturing labor and less energy are needed, leading to less contribution of population change to GHG emissions. While for US, its de-industrialization resulted in decreasing need for inputs, including labor force, in energy-intensive manufacturing industry, hence less contribution to emissions is made by population change.

6.2.4 Consumption Change

In this paper, driving factors for GHG emissions change are not only studied in production process, but also in consumption. Specifically, according to the decomposition results, changes in consumption can be split into two parts, namely consumption volume and pattern change.

As the consumption volume is expressed by final demand per capita, it reflects the worldwide demand for products in every country’s (China, India and US in this paper) sectors. Figure 22 compares the contributions of consumption volume change to GHG emissions in three countries from 1995 to 2009, where we can observe that such change in India led to about 25% increase of GHG emission by 2009 while similar contributions in China and US.

Further, figure 23 shows the change of consumption volume (in times) between the year 2009 and 1995 in three countries. As we can see, there are generally dramatic increase in consumption volume for China, for example, the consumption volume in sector 27 (Post and Telecommunications) in 2009

was about 43 times larger than that in 1995, it means, over the world, every person’s demand for Chinese Post and Telecommunications industry increased by 43 times in comparison to that in 1995 on average. Volume changes in India seems fewer than those in China and the least for US. But how should we understand the seemingly different patterns in two figures?

In this part, I have to admit the limit in figure 23, that is all the numbers for final demand in such figure are in current price term, which means that there are sizable price effects in comparing the consumption volume change between 1995 and 2009 for three countries. Such effects are not neglectable since Chinese economy developed so well during its industrialization on the one hand, with huge growth in national income and the capacity of export, thus leading to great price changes. On the other hand, US as one of the most developed countries, it always plays a role of price maker, thus price changes have little impact on US. Therefore, it is reasonable to believe that the consumption volume change in China may not be larger than that in India if the price effects are eliminated, leading to the results that are consistent with figure 22.

Differently, better results are shown by consumption pattern change which reflects how much is needed by the world for every sector in certain country. Figure 24 shows the contributions of such change to GHG emissions in three countries during 15 years. There are three curves depict different trends, among which consumption pattern change induced most GHG emissions in China, followed by the contribution in India while such change in US resulted in mitigation of GHG emissions. To further understand the three different patterns, we should firstly know about how did the consumption pattern change in three countries from 1995 to 2009.

Figure 25 illustrates the consumption pattern change in three countries best. The same as figure 23, data that applied in this figure also contains price effects, however, the weight of each sector’s consumption is calculated as the aggregated final demand by the whole world for each sector in a country divided by the world demand for all sectors in this country. Since the two terms above both include price effects, the ratio of them makes the effects less impactive. As the figure shows, during 1995 and 2009 in China, more energy-intensive products were demanded in several manufacturing sectors such as Machinery (13), Electrical and Optical Equipment (14) and Transport Equipment (15) comparing to increased consumption in service sectors in India. In US, there was even more overall decrease in demand of manufacturing products, leading to less energy use. Phenomena as such are understandable since the industrialization in China improved the productivity in manufacturing industries, together with the huge amount of cheap labor, Chinese manufactured products have the comparative advantage over those in other countries, leading to more export and consumption, thus inducing more GHG emissions. While for India, due to its failure in initiating industrialization, it is less likely to have more manufactured goods demanded by the world, instead, more consumption happened in tourism related service sectors. More clearly, for US, as it had already realized industrialization decades ago and other countries were emerging, especially China, it gradually lost the advantage in producing manufactured goods and instead, it imported such goods from other emerging countries. With the de-industrialization happened in US, its industrial structure tilted towards service sectors which are generally less energy-intensive, US provided the world with more service products, contributing to the mitigation of GHG emissions.

Conclusions

which is widely applied by previous works, namely emission coefficient (k), structural effects (M), population (p), consumption volume (yv) and consumption pattern (ys), with direct changes in

households emissions.

According to the results, it is the same for three countries that the world population change contributed most to increasing their GHG emissions, followed by less contribution made by global consumption volume change and national households emissions while global technological change commonly had opposite effect on their GHG emissions. Differently, structural change that happened over the world had positive effects on mitigation of GHG emissions in US and India but resulted in more emissions in China. Moreover, changes in world consumption pattern contributed to more emissions in China and India, but fewer emissions in US. However, even though some factors had similar contribution trends in three countries, statistical distinctions among which also reveal the truth that different stages of industrialization made them happen.

In the rest parts of analysis, drivers of GHG emissions in three countries are discussed individually. By doing so, differences of the same driver for three countries are able to reflect countries’ national changes. In that part, I provided detailed evidence for different consequences caused by different industrialization stages in three countries, for example, contributions of structural change to GHG emissions are further decomposed at sectoral level, the results prove that the de-industrialization in US induced mitigation of GHG emissions in both agriculture and manufacturing but more emissions in service sectors while premature de-industrialization resulted in more emissions in not only service sectors but also a minority of manufacturing industries, left mitigation of emissions in the rest of manufacturing industries. Nevertheless, Chinese industrialization led to more emissions concentrating on manufacturing industries. Similarly, explanations for other factors are also consistent with the characteristics of different stages of industrialization in US, China and India.

References

1. Bradford D F, Schlieckert R, Shore S H. The environmental Kuznets curve: exploring a fresh specification[R]. National Bureau of Economic Research, 2000.

2. Brodie F J. Decrease of fog in London during recent years[J]. Quarterly Journal of the Royal Meteorological Society, 1905, 31(133): 15-28.

3. Clay K, Troesken W. Did Frederick Brodie Discover the World's First Environmental Kuznets Curve? Coal Smoke and the Rise and Fall of the London Fog[R]. National Bureau of Economic Research, 2010.

4. Clark C. The conditions of economic progress[J]. The conditions of economic progress., 1967. Cairncross, A (1982) 'What is deindustrialisation?' Pp. 5–17 in: Blackaby, F (Ed.) Deindustrialisation, London: Pergamon2)

5. Cardarelli M R, Lusinyan L. US Total Factor Productivity Slowdown: Evidence from the US States[M]. International Monetary Fund, 2015.

6. Das D, Erumban A, Aggarwal S, et al. Total factor productivity growth in India in the reform period: A disaggregated sectoral analysis[C]//World KLEMS Conference, Harvard University. 2010.

7. Dietzenbacher E, Los B. Structural decomposition techniques: sense and sensitivity[J]. Economic Systems Research, 1998, 10(4): 307-324.

8. Feng K, Davis S J, Sun L, et al. Drivers of the US CO2 emissions 1997-2013[J]. Nature communications, 2015, 6.

9. Gordon R J. Is US economic growth over? Faltering innovation confronts the six headwinds[R]. National Bureau of Economic Research, 2012.

10. Gordon R J. US productivity Growth: The Slowdown has returned after a temporary revival[J]. International Productivity Monitor, 2013 (25): 13.

11. Grossman G M, Krueger A B. Environmental impacts of a North American free trade agreement[R]. National Bureau of Economic Research, 1991.

12. Grossman G M, Krueger A B. Economic growth and the environment[R]. National Bureau of Economic Research, 1994.

14. Hoekstra R, Van Den Bergh J C J M. Structural decomposition analysis of physical flows in the economy[J]. Environmental and resource economics, 2002, 23(3): 357-378.

15. Lever, W F (1991) ‘Deindustrialisation and the Reality of the Post-industrial City’ Urban Studies, Vol. 28, No. 6, Pp. 983-999

16. Lawrence R Z, Edwards L. US Employment deindustrialization: insights from history and the international experience[J]. Policy Brief, 2013 (13-27).

17. Lee C C, Chiu Y B, Sun C H. Does one size fit all? A reexamination of the environmental Kuznets curve using the dynamic panel data approach[J]. Applied Economic Perspectives and Policy, 2009, 31(4): 751-778.

18. Leontief W. Environmental repercussions and the economic structure: an input-output approach[J]. The review of economics and statistics, 1970: 262-271.

19. Marsiglio S, Ansuategi A, Gallastegui M C. The environmental Kuznets curve and the structural change hypothesis[J]. Environmental and Resource Economics, 2015: 1-24.

20. Nagdeve D A. Environment and health in India[C]//International Institute for Population Sciences, IUSSP Regional Population Conference on'Southeast Asia's Population in a Changing Asian Context at Bangkok, Thailand. 2002, 1013.

21. Ozyurt S. Total factor productivity growth in Chinese industry: 1952–2005[J]. Oxford Development Studies, 2009, 37(1): 1-17.

22. Panayotou T, Peterson A, Sachs J D. Is the Environmental Kuznets Curve driven by structural change? What extended time series may imply for developing countries[J]. 2000.

23. Panayotou T. Empirical tests and policy analysis of environmental degradation at different stages of economic development[R]. International Labour Organization, 1993.

24. Rodrik D. Premature deindustrialization[J]. Journal of Economic Growth, 2015: 1-33.

25. Shafik N, Bandyopadhyay S. Economic growth and environmental quality: time-series and cross-country evidence[M]. World Bank Publications, 1992.

Appendix

Figure 2: N-shaped EKC Source: Marsiglio at el (2015) Figure 1: Environmental Kuznets Curve Source: Panayotou (1993)

Figure 4: US manufacturing employment, 1990–2011 Source: Bureau of Labor Statistics

high-speed growth in industry

imbalanced structure

slow down the growth in industry

re-balanced structure

Figure 5: Cyclic Process in Chinese Industrialization

Agriculture Manufacturing Services

Figure 9: Employment share in India (in %)

Source: International Labor Organization, World Bank, MRI Figure 7: India GDP Growth in History (billion USD)

Source: www. tradingeconomics.com

Figure 11: Contributions of Different Factors to GHG Emissions in China (in ratio), 1995-2009

Figure 13: Contributions of Structural Change to GHG Emissions (in ratio), 1995-2009

Figure 14: Sectoral Contributions of Structural Change to GHG Emissions in US (in %), 1995-2009

Figure 17: Contributions of Technological Change to GHG Emissions (in ratio), 1995-2009

Figure 16: Sectoral Contributions of Structural Change to GHG Emissions in China (in %), 1995-2009

Figure 21: Manufacturing Employment Share in US (% of total employment), 1995-2009 Source: The World Bank

Figure 19: National Population Changes (in %), 1995-2009

Figure 23: Consumption Volume Change between 1995 and 2009 (in times)

Figure 22: Contributions of Consumption Volume Change to GHG Emissions (in ratio), 1995-2009

1 Agriculture, Hunting, Forestry and Fishing 2 Mining and Quarrying

3 Food, Beverages and Tobacco 4 Textiles and Textile Products 5 Leather, Leather and Footwear 6 Wood and Products of Wood and Cork 7 Pulp, Paper, Paper , Printing and Publishing 8 Coke, Refined Petroleum and Nuclear Fuel 9 Chemicals and Chemical Products 10 Rubber and Plastics

11 Other Non-Metallic Mineral 12 Basic Metals and Fabricated Metal 13 Machinery, Nec

14 Electrical and Optical Equipment 15 Transport Equipment

16 Manufacturing, Nec; Recycling 17 Electricity, Gas and Water Supply 18 Construction

Figure 25: Consumption Pattern Change between 2009 and 1995 (in %)

Table 1. Numbers of Columns and Rows for Each Matrix

19 Sale, Maintenance and Repair of Motor Vehicles and Motorcycles; Retail Sale of Fuel 20 Wholesale Trade and Commission Trade, Except of Motor Vehicles and Motorcycles 21 Retail Trade, Except of Motor Vehicles and Motorcycles; Repair of Household Goods 22 Hotels and Restaurants

23 Inland Transport 24 Water Transport 25 Air Transport

26 Other Supporting and Auxiliary Transport Activities; Activities of Travel Agencies 27 Post and Telecommunications

28 Financial Intermediation 29 Real Estate Activities

30 Renting of M&Eq and Other Business Activities 31 Public Admin and Defence; Compulsory Social Security 32 Education

33 Health and Social Work