Analyzing the Driving Forces behind CO₂ Emissions in Energy-Resource-Poor and Fossil-Fuel-Centered Economies: Case Studies from Taiwan, Japan, and South Korea

Abstract

Based on the strong similarities between energy-resource-poor and fossil-fuel-centered economies (e.g., Taiwan, Japan, and South Korea) in terms of economy, culture, and energy usage characteristics, they should be analyzed collectively. This study adopted two-tier input-output structural decomposition analysis to identify the driving forces behind CO₂ emissions to these countries to the formulation of effective environmental policy. Data from the World Input-Output Database was used to decompose relative changes in CO₂ emissions into a range of technological advances, factor substitution, and final demand effects. Technological advances in energy (direct) contributed to a 77% reduction in Taiwan and a 34% reduction in South Korea. This is a clear indication that improving energy efficiency via technological advances should be a priority. In Japan in particular, there was a 22% reduction in CO₂ emissions attributable to technological advances in materials; hence, it is recommended that Taiwan and South Korea work to extensively develop eco-industrial parks to create industry clusters to promote resource/energy efficiency and reductions in CO₂ emissions. Decomposition results based on factor substitution revealed that a variety of strategies will be required, such as switching to fuels that are less carbon intensive, promoting the adoption of renewable energies, and implementing clean-coal technologies.

1. Introduction

The effects of climate change can be observed across the globe, and there is compelling scientific evidence indicating that most of the effects are anthropogenic (i.e., related to human activity). The accumulation of greenhouse gases in the atmosphere due to the burning of fossil fuels is a prime culprit in elevated surface temperatures. This in turn leads to impacts such as shifting seasons, rising sea-levels, the disappearance of Arctic sea-ice, and heat waves of higher intensity. Taiwan, Japan, and South Korea lack indigenous energy resources and therefore depend on imported fossil fuels. Fossil fuels (i.e., coal, oil, and natural gas) make up approximately 75 percent of the total primary energy supply in Japan and South Korea and 90 percent in Taiwan [1], resulting in high CO₂ emissions. In 2018, Japan, South Korea, and Taiwan were respectively responsible for 3.2%, 1.8%, and 0.8% of the world’s CO₂ emissions, which ranks them as the 5th, 7th, and 21st largest emitters of CO₂ in the world [2]. Note that the per capita emission of CO₂ in Japan, Taiwan, and South Korea respectively amounted 8.6, 10.8 and 11.7, which is far higher than the global average (4.4). In addition, all three of these countries depend on export-oriented industrial sectors. Thus, these countries face international pressure (through formal or informal diplomatic channels), which could have a profound impact on trade (e.g., carbon border adjustment mechanism). It should also be noted that these countries have extensive coastlines, such that any rise in sea level due to climate change could have serious economic consequences. Reducing CO₂ emissions is an important issue in these countries.

Changes in CO₂ emissions are influenced by many factors. Identifying the driving forces behind CO₂ emissions is crucial to the development of effective environmental policy. Researchers have yet to propose a quantitative framework by which to analyse the factors underlying changes in CO₂ emissions within groups of economies such as energy-resource-poor economies. Analysis frameworks have been developed to assess changes in CO₂ emissions in Taiwan [3], Japan [4], and South Korea [5] individually. Based on the strong similarities between these countries in terms of economy, culture, and energy usage characteristics, we posit that they could be treated collectively.

When designing or revising carbon mitigation policies, understanding the dynamics and underlying factors of CO₂ emissions is of primary importance. Input-output structural decomposition analysis (I-O SDA) is used to analyse economic changes by examining a set of static changes in key parameters within an input-output table. This approach has been widely applied in the energy sector to quantify the factors affecting changes in economic variables [6]. The I-O SDA technique differentiates between the effects of final demand, factor substitution, and technological advances, resulting in a clear picture of the relative importance of each underlying factor. This study adopted a two-tier KLEM (KLEM is an abbreviation for capital (K), labour (L), energy (E), and materials (M)) I-O SDA to enable the decomposition of factors from the perspective of inter-industry economic activity. To ensure consistency across countries, this study employed the World Input-Output Database (WIOD) published by the Statistical Office of the European Communities (Eurostat) as its main data source.

Essentially, a quantitative framework was developed by which to analyse the driving forces behind energy consumption and CO₂ emissions in specific energy-resource-poor economies. Cross-economy comparisons made it possible to identify distinctive characteristics that are common to these countries to guide the formulation of effective carbon reduction policies.

The remainder of this paper is structured as follows. The following section presents a review of the literature pertaining to decomposition analysis. Section 3 describes the methodology employed and the data used in this paper. Section 4 presents an overview of changes in energy consumption and the decomposition results in these energy-resource-poor economies. The final two sections respectively provide a discussion of policy implications and our conclusions.

2. Literature Review

Decomposition analysis has been widely used to examine the driving forces underlying changes within aggregated indicators over time. One strand of literature uses the production function with various inputs to decompose the dynamics of energy consumption or emissions [7]. In production-theoretical decomposition analysis (PDA), general production processes are modelled by constructing a best-practice frontier in conjunction with distance functions [7,8]. PDA is meant to reveal the effects of production efficiency and technology on energy consumption or emissions [9]. Note however that the results are often influenced by the number of entities considered [10].

Many studies have employed index decomposition analysis (IDA) to explore variations in energy consumption or CO₂ emissions [11,12,13,14,15,16,17]. Researchers have developed a number of indexing methods (e.g., Laspeyres, Paasche, Divisia, Fisher approaches, and Marshall-Edgeworth) and variations using different weights or base years [18,19]. Despite differences in mathematical specifications, the basic idea remains the same: changes in CO₂ emissions (additive or multiplicative) can be decomposed into factors associated with fluctuations in activity, structure, intensity, energy-mix, and emission-factor effects [20]. In recent years, IDA has gained favour due to the growing availability of data and its flexibility in applying disaggregation at different levels to facilitate empirical analysis and comparisons [21].

Another branch of the literature has focused on SDA (otherwise referred to as I-O SDA) to take advantage of the fact the underlying data structure includes economic activities between industries. The inclusion of I-O tables in the SDA framework makes it possible to conduct a thorough examination of technological advances and factor substitution (on the supply side) as well as changes in exports and final demand (on the demand side). Note that the I-O SDA model accounts for the effects of indirect demand (Indirect effects occur in situations where an increase in direct demand in one sector leads to an increase in the demand for inputs from other sectors), whereas IDA can only be used to assess direct effects. Overall, I-O SDA makes it possible to differentiate between a number of technological advances, factor substitution, and final demand effects, which are simply beyond the scope of the IDA framework. Note also that I-O tables are developed for economies of broader scope (e.g., at the national scale), whereas IDA studies tend to focus on one sector at a time [22,23]. The SDA method was therefore adopted for the current study.

SDA has been applied extensively in the field of energy and the environment, particularly in elucidating the driving forces behind energy consumption and CO₂ emissions. Applications of SDA can be found in studies on Denmark [24], Netherlands [25], Japan [4], China [26,27,28,29], Taiwan [30], Korea [5], Brazil [31], the U.S. [32], the UK [33], Portugal [34], Australia [35], and Thailand [36]. Note however that most previous SDA studies utilized only the one-tier (one-stage) I-O SDA model, which disregards various sub-effects from changes in input technology. This study sought to fill this gap in the literature by constructing a more complex two-tier I-O SDA model, based on the well-established KLEM production function involving capital (K), labour (L), energy (E), and material (M). In the first tier, all of the inputs are grouped into these four aggregate factors. In the second tier, energy (E) and material (M) (The two-tier KLEM production function can be expressed as: Q = f [K, L, E(E1, E2, …), M(M1, M2, …)]) are composed of different elements such as natural gas and oil in energy aggregate. Note that very few I-O SDA models enable the disaggregation of household consumption. This study sought to disaggregate household consumption into autonomous and induced consumption. The novelty of our SDA model is based on the well-established production function form (i.e., KLEM production function) and the well-accepted macro-economic theory (i.e., simple Keynesian theory).

Due to the increasing international fragmentation of production, sectors within and across economies are closely connected to each other through trade networks, leading to many influencing impacts across borders, such as shared gains for domestic factors, technological advances, factor substitution, and production specialization [37,38]. SDA differentiates between intermediate production and final demand within or between entities (e.g., sectors and economies) within the context of the global economy [39]. Nonetheless, the most important issue for this analysis is the consistency of data sources for the various countries. A number of studies have applied decomposition analysis to the WIOD to explore factors affecting changes in energy consumption or CO₂ emissions [40,41,42,43]. For example, Andreoni and Galmarini [44] and Voigt et al. [45] used IDA to investigate fluctuations in CO₂ emissions and energy intensity in multiple countries. Jiang and Guan [46] used SDA to study changes in CO₂ emissions in 40 countries between 1995 and 2009. Overall, the results of these studies revealed that in developing countries, demand for electricity drives the demand for coal, leading to enormous increases in CO₂ emissions. In developed countries, the demand for natural gas in the public sector and chemical industry also increased CO₂ emissions. Zhong [39] applied SDA to WIOD data to decompose relative changes in energy consumption into six major factors. They discovered that growth in energy consumption is driven mainly by final demand; however, they also determined that growth can be mediated by the intensity effect resulting from technological advances. Most previous research using decomposition analysis at the multinational or global scale has focused on IDA or one-tier SDA. These methods have proven effective in aggregative decomposition; however, they are ill-suited to exploring the complex combinations of driving forces underlying CO₂ emissions. The two-tier KLEM I-O SDA employed in the current study accounts for sub-effects due to changes in technology and provides more detailed information pertaining to decomposition factors.

The contributions of this paper are as follows:

• A two-tier KLEM I-O SDA framework was employed to investigate the driving forces underlying CO₂ emissions in a group of energy-resource-poor economies.
• Previous SDA studies typically analyzed source of CO₂ emissions limited to changes in final demand and broad groupings of technological advances. In contrast, the proposed model identifies various types of demand shift, induced changes in household consumption, factor substitution, technological advances, and import substitution. The proposed scheme provides a more comprehensive understanding of the relative importance of factors underlying changes in CO₂ emissions.

3. Method and Data

3.1. Two-Tier KLEM I-O SDA

According to the simple Keynesian model [47], household consumption can be divided into autonomous and induced consumption. Autonomous consumption is exogenous to the economy, whereas induced consumption is endogenous. In the current study, only induced consumption is included in the inter-industry transactions matrix, such that autonomous consumption remains in the final demand column as an exogenous variable. Wages in the primary input matrix are incorporated within the inter-industry transaction matrix, as the row corresponding to induced household consumption. Wages are therefore treated as a production factor by which to derive the level of induced household consumption as the product of marginal propensity to consume and total wages paid. The level of autonomous household consumption is derived by subtracting induced household consumption from the total household consumption.

Figure 1 illustrates hierarchical structure of the proposed I-O SDA model. In the first tier, changes in CO₂ emissions are distinguished by three major effects: (1) changes in induced demand coefficients, (2) import substitution, and (3) changes in autonomous final demand. These major effects can be further divided into sub-effects. For example, changes in induced demand coefficients can be divided into changes in industry production functions and changes in household induced consumption. Moreover, sub-effects can be decomposed once again, such that changes in industry production functions is split into technological advances and factor substitution.

Equations (1)–(6) present the decomposition equations used to examine changes in energy and CO₂ emissions. Our two-tier KLEM I-O SDA decomposes changes in energy and CO₂ emissions into six main effects: changes in exports (Equation (1)), changes in autonomous domestic final demand (Equation (2)), changes in household-induced consumption (Equation (3)), import substitution (Equation (4)), factor substitution (Equation (5)), and technological advances (Equation (6)). For detailed derivations of these equations, please refer to Wu et al. [48]. Essentially, the model involves a number of comparative static computations, which differ from one another only in terms of one set of key I-O coefficients. For example, in Equation (1), the effect of changes in exports on CO₂ emissions is modeled by the CO₂ emission factor (c), energy coefficient (B), induced demand coefficients (A^c), and exports (Ex). Equation (6) isolates technological changes within four constituent components: capital, labor, energy, and materials (KLEM). Thus, technological advances incorporate various sub-effects, including advances in energy (direct), advances in energy (linkage), advances in capital, advances in labor, and advances in materials.

$$c_{t - 1}{B}_{t - 1}{\left(I - R_{t - 1}{A}_{t - 1}^C\right)}^{-1}(E{x}_t - E{x}_{t - 1})$$

(1)

$$c_{t - 1}{B}_{t - 1}{\left(I - R_{t - 1}{A}_{t - 1}^C\right)}^{-1}\left[R_{t - 1}\left(Y_t^A - Y_{t - 1}^A\right) + E{x}_t - E{x}_{t - 1}\right]$$

(2)

$$c_{t - 1}{B}_{t - 1}\left\{{[I - R_{t - 1}(A_{t - 1}^P|b_t)]}^{-1}-{[I - R_{t - 1 }(A_{t - 1}^P|b_t)]}^{-1}\right\}\left[R_{t - 1}{Y}_t^{Ad} + E{x}_t\right]$$

(3)

$$c_t{B}_t\left[{\left(I - R_t{A}_t^C\right)}^{-1} - {\left(I - R_{t - 1}{A}_t^C\right)}^{-1}\right]\left[\left(R_t - R_{t - 1}\right)Y_t^{Ad} + E{x}_t\right]$$

(4)

$$\left\{c_t{B}_t{[I - R_{t - 1}(A_t^P|b_t)]}^{-1} - B_{t\left(t - 1\right)}{[I - R_{t - 1 }(A_{t\left(t - 1\right)}^P|b_t)]}^{-1}\right\}\left[R_{t - 1}{Y}_t^{Ad} + E{x}_t\right]$$

(5)

$$\left\{c_{t - 1}{B}_{t\left(t - 1\right)}{[I-R_{t - 1}(A_{t\left(t - 1\right)}^P|b_t)]}^{-1}-B_{t - 1}{[I - R_{t - 1 }(A_{t - 1}^P|b_t)]}^{-1}\right\}\left[R_{t - 1}{Y}_t^{Ad} + E{x}_t\right]$$

(6)

where superscript t and subscript t − 1 respectively refer to the year t and year t − 1; c is matrix of CO₂ emission factors related to various forms of energy; B is matrix of energy consumption per unit of outputs; I represents unit matrix; M denotes diagnose matrix of import coefficients; R means (I−M); Ex signifies column vector of exports; A^C = [A^P|b] indicates matrix of induced demand coefficients; A^P is matrix of industry production function; b is household induced consumption; Y^Ad is matrix of autonomous final demand. Change in induced demand coefficients can be decomposed into changes in household induced consumption (b), and changes in industry production functions (A^P). For the sake of clarity, A^C is written as: A^C = [A^P|b]. The notation does not involve calculation of any matrices. Instead, the notation simply indicates that change in matrix A^C can be attributed to changes in matrices A^P and b.

The concept of decomposition in the two-tier KLEM I-O SDA is similar to the Laspeyres and Paasche index in terms of constant weights, which means that when a given factor changes, the other factors remain fixed during the same base period. Unfortunately, this practice tends to produce residual terms, which can make interpretation difficult. The adoption of variable weights for decomposition could eliminate residual terms; however, this would initiate path dependence, wherein the selection of paths could influence the final result. Thus, this study adopted the weighted Shapley value proposed by Albrecht [49] to overcome the shortcomings of the two-tier KLEM I-O SDA [3]. The weighted Shapley value was employed to construct a generalized equation by which to generate a set of solutions to I-O SDA and verify that they can avoid the need for arbitrary determinations involving ad hoc decomposition forms, provide a solution to the non-uniqueness problem associated with decomposition, and eliminate residual terms.

3.2. Data Sources

The WIOD includes national I-O tables, socioeconomic accounts (SEAs), and environmental accounts. SEAs contain industry-level data on employment, capital stocks, gross output, and value added, whereas environmental accounts contain industry-level data pertaining to energy use and CO₂ emissions. The first edition of these datasets was released in 2013. The System of National Accounts 1993 published by the United Nations (UN) provides datasets covering 40 countries classified into 35 sectors in accordance with the International Standard Industrial Classification Revision 3 (ISIC Rev. 3). The time periods for these datasets vary from 1995 to 2011 [50]. A new edition of the I-O tables was released in 2016 covering the period from 2000 to 2014; however, energy and environmental data tables have yet to be updated. The newest edition of the database was published by the System of National Accounts 2008, covering 43 countries classified into 56 sectors in accordance with the International Standard Industrial Classification Revision 4 (ISIC Rev. 4).

The period covered by environmental accounts extended only to 2009; therefore, the tables published by Corsatea et al. [51] were adopted, which are consistent with WIOD 2016, including gross energy use, emission relevant energy use, and CO₂ emissions pertaining to 12 energy commodities in 64 sectors. The data covers 28 European Union (EU) countries and 13 other major countries from 2000 to 2016. The sectors in the updated environmental accounts were inconsistent with the I-O tables; therefore, this study implemented sectoral classification and aggregation. Furthermore, the national I-O tables associated with WIOD 2016 extend only to 2014; therefore, this study was limited to data covering the decade prior to publication (i.e., 2004–2014). Other data, such as labour compensation, capital compensation, nominal gross fixed capital formation, price levels of gross output, price levels of intermediate inputs, price levels of gross value added, and price levels of gross fixed capital formation, were obtained from SEAs. The coefficients of energy-related CO₂ emissions were derived from Intergovernmental Panel on Climate Change (IPCC) guidelines [52].

National I-O tables are based on monetary units. It is necessary therefore to deflate the I-O table to a constant monetary value to avoid influence from annual fluctuations in price levels. This study rebalanced the national I-O tables using various SEA price indices and a bi-proportional matrix routine similar to the RAS (RAS is a widely-used approach to balancing I-O tables. This approach introduced by Richard A. Stone reduces rounding errors in computation of the Rs and Ss, which represent the series of sequential row and column modifications, i.e., diagonal matrices) method [53]. To avoid double counting final energy consumption, selected data related to the transformation of energy in environmental accounts were not included in our model, including (1) energy sources for power generation and (2) coking coal, coke, and intermediate products (e.g., furnace gas) for steel making. Finally, calculation of CO₂ emission coefficients of various energy sources in the WIOD was based on the weighted average of the emission coefficients from IPCC guidelines.

4. Results

4.1. Changes in Energy Consumption in Energy-Resource-Poor and Fossil-Fuel-Centered Economies

4.1.1. Changes in Energy Consumption in Taiwan: Major Industries and Overall

Between 2004 and 2014, CO₂ emissions in Taiwan increased by 7%, most of which can be attributed to the burning of fossil fuels (e.g., coal, oil, and natural gas). During that period, coal consumption increased by 4%, oil consumption decreased by 19%, gas consumption increased by 74%, and electricity consumption increased by 23%. This study further scrutinized changes in energy consumption in major industries in order to clarify the relationship between changes in CO₂ emissions and industrial economic activity (see Figure 2). In Taiwan, the dominant industry contributing to the growth of coal consumption was the manufacture of chemicals and chemical products, whereas the main industries driving the decrease in oil consumption were the electricity, gas, steam, and air-conditioning supply sectors, as well as the manufacture of chemicals and chemical products. The primary industries responsible for the increase of gas consumption were the electricity, gas, steam, and air-conditioning supply sectors. The increase in electricity consumption in Taiwan can be attributed primarily to export-oriented industries (e.g., the manufacture of electronic products), leading to an increase in overall CO₂ emissions.

4.1.2. Changes in Energy Consumption in Japan: Major Industries and Overall

Between 2004 and 2014, CO₂ emissions in Japan increased by 1%. Specifically, coal consumption increased by 6%, oil consumption decreased by 24%, gas consumption increased by 46%, and electricity consumption decreased by 4%. The dominant industries contributing to the increase in coal consumption was the electricity, gas, steam, and air-conditioning supply sectors, whereas the main industries driving the increase in gas consumption were the electricity, gas, steam, and air-conditioning supply sectors, education, the refining of basic metals, the manufacture of rubber and plastic products, and the manufacture of food products. No significant changes were observed in electricity consumption due to mutual offsets between industries. Overall, there was a decline in oil consumption in demand-oriented sectors in Japan (e.g., land transport and transport via pipelines, crop and animal production, hunting, and wholesale trade). The decreases in coal consumption and oil consumption in export-oriented industries (e.g., the refining of basic metals and the manufacture of chemicals and chemical products) played an important role in curbing the increase in total CO₂ emissions.

4.1.3. Changes in Energy Consumption in South Korea: Major Industries and Overall

Between 2004 and 2014, CO₂ emissions in South Korea increased by 30%. Specifically, coal consumption increased by 45%, oil consumption decreased by 19%, gas consumption increased by 84%, and electricity consumption increased by 40%. The main industry driving the increase in coal consumption was the electricity, gas, steam, and air-conditioning supply sectors. The primary industries contributing to the growth of gas consumption were the electricity, gas, steam, and air-conditioning supply sectors, the manufacture of chemicals and chemical products, and the refining of basic metals. The dominant industries responsible for the increase in electricity consumption were the refining of basic metals, the manufacture of electronic products, and water collection, treatment, and supply. In summary, the increase in electricity consumption in export-oriented industries (e.g., the refining of basic metals and the manufacture of electronic products) and demand-oriented industries (e.g., electricity, gas, steam and air-conditioning supply) were the main factors driving the increase in total CO₂ emissions.

4.2. Decomposition Results

4.2.1. Decomposition Results: Taiwan

Figure 3 presents results from decomposition analysis, illustrating the means by which the determinants influenced CO₂ emissions in these energy-resource-poor and fossil-fuel-centered economies. In Taiwan, the increase in CO₂ emissions during 2004–2014 can be attributed primarily to an increase in exports, reflecting the manufacturing-driven and export-oriented characteristics of the Taiwanese economy. Increased exports drove up CO₂ emissions due to coal consumption (+62%), oil consumption (+90%), gas consumption (+70%), and electricity consumption (+71%), which resulted in a 72% increase in total CO₂ emissions (

Table 1. Sources of changes in energy and CO₂ emissions in Taiwan between 2004 and 2014.

Determinants (%) 1	Coal	Oil	Gas	Electricity	CO2
Changes in exports	62	90	70	71	72
Changes in autonomous domestic final demand	5	4	7	5	5
Changes in household-induced consumption	2	2	3	0 2	2
Import substitution	−17	−5	−1	1	−9
Factor substitution	−8	−38	47	−5	−11
Technological advances	−40	−72	−52	−49	−51
Total Changes	4	−19	74	23	7

¹ Values in the table are defined as [∆V/V0∗100%], where “V0” refers to the amount of energy and CO₂ emissions during the base year 2004 and “∆V” refers to the change in volumes incurred by each factor. ² Zero is the result of rounding. The actual value is −0.18%.

). The manufacture of chemicals and chemical products drove up coal consumption, whereas the electricity, gas, steam, and air-conditioning supply sectors drove up gas consumption. The manufacture of electronic products and the manufacture of chemicals and chemical products drove up electricity consumption.

Factor substitution drove up the consumption of gas by 47%, but reduced coal consumption by 8%, oil consumption by 38%, and electricity consumption by 5%. This trend toward replacing coal and oil with gas reduced total CO₂ emissions by 11%. Technological advances reduced CO₂ emissions due to coal consumption (−40%), oil consumption (−72%), gas consumption (−52%), and electricity consumption (−49%), which resulted in a 51% decrease in total CO₂ emissions. These findings demonstrate the degree to which technological advances have decreased energy usage per unit of output.

4.2.2. Decomposition Results: Japan

In Japan, technological advances led to a 15% decrease in CO₂ emissions between 2004 and 2014. Factor substitution led to a 15% increase in CO₂ emissions and increased exports led to a 7% increase (

Table 2. Sources of changes in energy and CO₂ emissions in Japan between 2004 and 2014.

Determinants (%)	Coal	Oil	Gas	Electricity	CO2
Changes in exports	12	1	9	8	7
Changes in autonomous domestic final demand	1	1	4	1	2
Changes in household-induced consumption 1	0	0	0	0	0
Import substitution	−14	−3	−8	−7	−9
Factor substitution	15	2	48	10	15
Technological advances	−8	−25	−7	−16	−15
Total Changes	6	−24	46	−4	1

¹ Zeros are the result of rounding. The actual values are 0.007%, −0.016%, 0.028%, −0.025%, and −0.003%, respectively.

). Technological advances reduced CO₂ emissions due to coal consumption (−8%), oil consumption (−25%), gas consumption (−7%), and electricity consumption (−16%), which resulted in a 15% decrease in total CO₂ emissions.

Import substitution reduced coal consumption (−14%), oil consumption (−3%), gas consumption (−8%), and electricity consumption (−7%), which resulted in a 9% decrease in total CO₂ emissions. This can be attributed primarily to an increase in the import coefficient in most industries. Changes in autonomous domestic final demand led to increases in coal consumption (+1%), oil consumption (+1%), gas consumption (+4%), and electricity consumption (+1%), which resulted a 2% increase in total CO₂ emissions. This indicates that the increase in autonomous domestic final demand slightly increased CO₂ emissions in Japan.

4.2.3. Decomposition Results: South Korea

In Korea, changes in exports increased CO₂ emissions by 30% and changes in autonomous domestic final demand increased CO₂ emissions by 24% (

Table 3. Sources of changes in energy and CO₂ emissions in South Korea between 2004 and 2014.

Determinants (%)	Coal	Oil	Gas	Electricity	CO2
Changes in exports	30	26	35	34	30
Changes in autonomous domestic final demand	28	16	32	24	24
Changes in household-induced consumption 1	1	0 1	1	0 2	0 3
Import substitution	−4	−3	1	0	−2
Factor substitution	16	−35	52	13	4
Technological advances	−27	−23	−37	−31	−28
Total Changes	45	−19	84	40	30

¹ Zero is the result of rounding; the actual value is −0.360%. ² Zero is the result of rounding; the actual value is 0.017%. ³ Zero is the result of rounding; the actual value is 0.481%.

). Increased exports drove up CO₂ emissions due to coal consumption (+30%), oil consumption (+26%), gas consumption (+35%), and electricity consumption (+34%), which resulted in a 30% increase in total CO₂ emissions. The manufacture of chemicals and chemical products and the refining of basic metals increased gas consumption, whereas the refining of basic metals and manufacture of electronic products increased electricity consumption, resulting in a 30% increase in total in CO₂ emissions.

Factor substitution decreased oil consumption by 35%, but increased coal consumption by 16%, gas consumption by 52%, and electricity consumption by 13%. This trend toward replacing oil with coal or natural gas resulted in a 4% increase in CO₂ emissions. Changes in autonomous domestic final demand drove up coal consumption (+28%), oil consumption (+16%), gas consumption (+32%), and electricity consumption (+24%). Land transport and transport via pipelines increased oil and gas consumption, whereas the electricity, gas, and water supply sectors drove up coal, gas, and electricity consumption, resulting in a 24% increase in CO₂ emissions.

5. Discussion

Our analysis of these energy-resource-poor and fossil-fuel-centered economies (Taiwan, Japan, and South Korea) revealed four determinants—changes in exports, technological advances, import substitution, and factor substitution—are with a profound effect on CO₂ emissions but different implications for three countries. Decomposition results for Taiwan and South Korea revealed that the increase in CO₂ emissions can be mainly attributed to increased exports. This is probably due to production expansion (e.g., manufactured goods) associated with increasing market demand. For example, continued growth in exports fuelled an expansion in the manufacture of electronic products, with a corresponding increase in electricity consumption and CO₂ emissions. Changes in the exports effect in Japan were less pronounced than those in Taiwan and South Korea, due primarily to a slowdown in the growth of export-oriented industries, which accounted for only 7% of the increase in CO₂ emissions.

Technological advances can be divided into sub-effects, including advances in energy (direct), advances in energy (linkage), advances in labour, and advances in materials. Technological advances were shown to contribute significantly to lowering total CO₂ emissions, accounting for 51% of the reduction in Taiwan, 15% in Japan, and 28% in South Korea; however, the three countries differed in terms of sub-effects (

Table 4. Influences of sub-effects of technological advances on CO₂ emissions in Taiwan, Japan, and South Korea between 2004 and 2014.

Determinants (%)	Taiwan	Japan	South Korea
Technological advances	−51	−15	−28
energy (direct)	−77	9	−34
energy (linkage)	1	−1	0.5
labour	5	−1	0.2
materials	20	−22	5

). Decomposition analysis revealed that technological advances in energy (direct) had the most pronounced effects in Taiwan and South Korea, whereas advances in materials made the greatest contribution in Japan. All these three countries improved their efficiency in the consumption of energy and utilization of materials during the study period. Japan is somewhat limited in the degree to which it can improve energy efficiency; however, advances in energy (direct) accounted for 77% of the decrease in CO₂ emissions in Taiwan and 34% in South Korea. This indicates that enhancing energy efficiency through technological advances should be a priority in energy-resource-poor and fossil-fuel-centered economies.

Technological advances in materials decreased CO₂ emissions by 22% in Japan, perhaps due to policies implemented by the Japanese government encouraging the development of high-value-added and low-carbon products. This is an indication that other countries with similar economies would benefit from switching to industrial processes with lower CO₂ emissions. Governments should also promote cooperation between different sectors. Eco-industrial parks comprising clusters of disparate industries have proven highly effective in reducing overall CO₂ emissions. This type of industrial symbiosis has been shown to minimize energy consumption, material usage, and environmental loads through the use of recycling networks for raw materials, waste, and energy.

Import substitution was shown to decrease CO₂ emissions in all three countries, which resulted in an increase in the import coefficients of most industries during the study period. This can be explained by the fact that domestic output in some industries was largely replaced by foreign products, probably due to lower production costs in countries such as China and Southeast Asia. Increases in import coefficients can also be attributed to the heavy dependence of all three economies on the electronics industry, which forms a global production chain. In this scenario, any form of industrial expansion will tend to increase the quantity of semi-finished goods that must be imported from abroad. This link between the import coefficient and CO₂ emissions is a clear indication of the degree to which trade patterns can affect environmental metrics.

Factor substitution among the various types of energy decreased total CO₂ emissions in Taiwan by 11%; however, it increased total CO₂ emissions by 15% in Japan and by 4% in South Korea. This can be attributed to the fact that the Taiwan government has been shutting down or scaling back outdated coal-fired power plants in favour of natural gas. If Japan and South Korea expect to achieve similar reductions, they will need to switch to less carbon-intensive power plants (Although coal-to-gas switching can reduce CO₂ emissions, it may increase the electricity costs). Renewable energy accounts for only a small proportion of the energy mix in these countries; however, administrative and financial measures could help promote renewable energy (Although renewable energy sources account for a negligible proportion of greenhouse gas emissions and other air pollution, some weaknesses and barriers associated with development of renewable energy should be acknowledged. Renewable energy sources (particularly electricity generation) depend entirely on geographical location and weather conditions. Current volatility and unpredictability in the renewable energy sector as well as various legal and social barriers (e.g., people’s attitude) hamper the adoption of renewable energy [54,55,56,57]. These barriers could probably be alleviated by educating end-users on the benefits of renewable energy, or by implementing legislation to support renewable energy) in order to reduce CO₂ emissions per unit of power generation.

6. Conclusions

This paper developed a quantitative framework by which to examine the issues associated with CO₂ emissions in three economies over a 10-year period. In the current study, economies characterized by poor energy resources and dependence on fossil fuels were examined; however, our use of a unified framework could be extended to any group of economies with common characteristics. A two-tier KLEM I-O SDA was employed based on the well-established KLEM production function involving capital (K), labour (L), energy (E), and material (M) as well as various sub-effects. Unlike the conventional IDA framework, I-O SDA differentiates among technological advances, factor substitution, and final demand effects. This makes I-O SDA an ideal tool to examine the driving forces behind CO₂ emissions, due to the fact that changes in data are more meaningful than the data itself in the formulation of policy. The proposed methodology could also be extended to the study of carbon emissions in other countries with common characteristics.

Our decomposition results pertaining to Taiwan, Japan, and South Korea reveal that the following four factors exert an overwhelming effect on CO₂ emissions: changes in exports, technological advances, import substitution, and factor substitution. Technological advances were shown to enhance efficiency in the consumption of energy and materials in all three countries. Advances in energy (direct) accounted for 77% of the decrease in CO₂ emissions in Taiwan and 34% in South Korea. This is an indication that improving energy efficiency through technological advances and innovation should be a policy priority in such economies. Technological advances in materials decreased CO₂ emissions by 22% in Japan. This is an indication that other countries with similar economies would benefit from switching to industrial processes with lower CO₂ emissions. This study therefore recommend that governments could promote cooperation between different sectors through the development of eco-industrial parks comprising clusters of disparate industries.

The fact that import substitution contributed to a decline in CO₂ emissions in all three countries demonstrates how resource-poor countries depend on trade and how trade patterns can affect CO₂ emissions. Factor substitution among the various types of energy decreased total CO₂ emissions in Taiwan by 11%; however, it increased total CO₂ emissions by 15% in Japan and by 4% in South Korea. Reducing CO₂ emissions will require the replacement of conventional coal-fired generators in favour of cleaner fuels (e.g., natural gas) and renewable energy technologies (e.g., wind power and solar energy) and clean-coal technologies.

Despite its valuable contributions, this paper is subject to a number of limitations and constraints. First, CO₂ emissions were examined within industrial sectors; i.e., we disregarded other types of greenhouse gas emissions and the entire household sector. Second, for the sake of consistency between countries, the WIOD was employed; however, we were able to obtain complete datasets only for the period leading up to 2014. Subsequent analysis using more up-to-date information remains contingent on updates by Eurostat or other research institutions.