Input–Output Analysis of Greenhouse Gas Emissions from Fossil Fuel Combustion Associated with Paper Products in Thailand: A Comparative Study of 2015 and 2021

2.1 Study framework and data sources

This study adopts an IO-based environmental assessment framework to quantify GHG emissions from fossil fuel combustion across economic sectors in Thailand, with a comparative analysis between 2015 and 2021 [6, 7]. The IO framework has been widely applied in environmental and energy studies to capture intersectoral dependencies and embodied emissions across supply chains, particularly in national-scale assessments [8, 9]. Unlike process-based approaches, IO models provide a comprehensive representation of indirect emissions, which are often underestimated in sector-specific analyses [5, 10]. Rather than relying on a fully integrated environmentally extended input–output (EEIO) database, a customized framework is developed by integrating national IO tables with sectoral energy consumption data and emission factors. This hybridized approach has been shown to improve consistency between economic and environmental datasets while maintaining transparency in emission estimation [7, 11].

The analytical procedure consists of four main steps: (i) estimating sectoral direct emissions based on fossil fuel consumption, (ii) conducting a comparative structural analysis to examine the key drivers of emission changes between 2015 and 2021, (iii) incorporating emission coefficients into the IO framework to estimate total emissions, and (iv) deriving emission intensity indicators.

Energy consumption data were obtained from Thailand Energy Efficiency Situation and Energy Balance reports for 2015 and 2021 [4, 12, 13]. Emission factors were sourced from TGO [14], and economic data were derived from Thailand’s IO tables compiled by the Office of the National Economic and Social Development Council [15].

To ensure comparability between the 2015 and 2021 monetary IO data, all monetary variables used for the temporal comparison were expressed at constant 2015 prices. The 2015 IO table was retained as the base-year price structure, while the 2021 monetary flows were deflated using the closest available producer price indices (PPI) from the Trade Policy and Strategy Office (TPSO), Ministry of Commerce, Thailand. The PPI series was selected according to the Classification of Products by Activity (CPA), with 2015 as the base year (2015 = 100) and 2021 as the target year. The price harmonization was applied consistently to final demand, intermediate transactions, gross output, and the corresponding IO coefficients used in the 2021 model.

Because the monetary IO tables for 2015 and 2021 are expressed in nominal/current prices, the 2021 IO data were converted to a common 2015 constant-price basis before conducting the temporal comparison. Sector-matched price deflation was performed using the closest available producer or service price indices from the TPSO, with 2015 = 100 as the reference price basis. The price harmonization was applied consistently to intermediate transactions, total output, and final demand. The technical coefficient matrix and Leontief inverse for 2021 were subsequently recalculated using the price-harmonized IO data.

2.2 System boundary and scope definition

The system boundary of this study is defined to include GHG emissions arising exclusively from fossil fuel combustion within domestic economic activities in Thailand. This definition follows established guidelines for energy-related emissions accounting and ensures consistency with international reporting standards [16, 17].

The analysis encompasses all economic sectors and fossil fuel types considered in this study, with detailed classification and data harmonization procedures provided in Section 2.3 to ensure consistency between energy consumption data and the IO framework.

Within this boundary, direct fossil-fuel combustion emissions are estimated for all 14 economic sectors based on sectoral fuel consumption data, including Public Utilities (PubU) and Transport and Communication (T&C). These sectoral emissions are subsequently propagated through intersectoral linkages in the domestic IO framework to capture indirect emissions associated with final demand. Accordingly, fossil-fuel combustion emissions from upstream electricity generation and transportation activities are included when they are represented as domestic sectoral production requirements in the IO system. Imported intermediate inputs and imported final goods are not separately modeled by foreign producing sectors; therefore, fossil-fuel emissions embodied in foreign upstream production are outside the system boundary. The resulting estimates represent emissions associated with domestic production requirements within Thailand rather than the full global supply chain [8, 16].

2.3 Data sources and integration

This study integrates three primary datasets: (i) sectoral energy consumption, (ii) emission factors, and (iii) economic IO tables, to construct a consistent analytical framework for estimating GHG emissions.

Energy consumption data were obtained from Thailand’s national energy statistics [4, 12, 13], which provide detailed fossil fuel consumption across sectors and fuel types. Emission factors were sourced from TGO [14], ensuring consistency with national GHG accounting practices. Economic data were derived from IO tables compiled by NESDC [15].

To ensure consistency between the economic and energy datasets, the original 16-sector classification of the IO tables was aggregated into the 14-sector classification used in the national energy statistics. Specifically, the “Other” and “Unclassified” sectors were combined into “Unclassified (UnCl)”, while the “Trade” and “Services” sectors were combined into “Trade and Services (T&S)”. This sector aggregation established a consistent sector classification between the IO tables and the national fuel consumption data.

Following this sector aggregation, fossil fuel consumption data were mapped to the corresponding IO sectors using a sector–fuel concordance. The fuel types include liquefied petroleum gas (LPG), kerosene, premium gasoline, regular gasoline, high-speed diesel, low-speed diesel, fuel oil, natural gas, and coal, consistent with the national energy statistics. The resulting sector–fuel mapping is summarized in Table 1, based on observed fuel consumption in 2015 and 2021 [4, 12, 13].

The integration of physical energy data with monetary IO tables is a critical step in environmentally extended IO analysis and has been widely applied in previous studies [7, 9, 18]. This approach ensures consistency between physical energy flows and economic transactions while maintaining methodological transparency.

For temporal comparison, the monetary IO data were harmonized to constant 2015 prices. The 2021 price adjustment used sector-matched price deflators derived from the closest available TPSO producer/service price indices classified according to the CPA. This treatment was necessary because directly corresponding price indices were not available for all aggregated IO sectors [19-21].

Table 1. Sector-fuel mapping framework used in this study based on the harmonized 14-sector classification

Note: ✓ indicates the presence of fuel consumption based on national energy statistics (2015 and 2021). LPG = liquefied petroleum gas; NG = natural gas.

2.4 Input–output model formulation and price harmonization

The intersectoral transaction matrix $Z$ is derived from the IO tables and represents the monetary flows between economic sectors [7]. This matrix forms the basis for describing the interdependencies of production activities within the economy and has been widely applied in economic and environmental system analysis [8, 22].

The technical coefficient matrix $A$ is calculated by normalizing intersectoral transactions by total sectoral output. Each element $A_{i j}$ represents the input required from sector $i$ to produce one unit of output in sector $j$, thereby reflecting the production structure of the economy [7, 23]. In practical terms, this matrix quantifies how production in one sector depends on inputs from other sectors and is fundamental to IO modeling.

The technical coefficient matrix is defined as:

$A_{i j}=\frac{Z_{i j}}{x_j}$   (1)

where, $Z_{i j}$ represents the monetary flow from sector $i$ to sector $j$, and $x_j$ is the total output of sector $j$ [7].

The total output required to satisfy final demand is expressed as:

$x=(I-A)^{-1} y$   (2)

where, $I$ is the identity matrix, and $y$ is the final demand vector. The Leontief inverse matrix $(I-A)^{-1}$ represents the total (direct and indirect) production requirements across all sectors and forms the core analytical component for tracing supply chain effects [7, 9, 22].

Price harmonization of monetary IO data: to enable a meaningful comparison between 2015 and 2021, the 2015 IO table was retained as the reference price basis, while the 2021 monetary transactions were converted to 2015 constant prices using sector-matched price indices from the TPSO. Let $P_i$ denote the 2021 price index for sector $i$, with 2015 = 100. The corresponding deflation factor was calculated as $P_i / 100$, and the 2021 monetary values were converted to 2015 constant prices by dividing the corresponding values by this factor. As shown in Table 2, sector-specific price indices were assigned to the aggregated IO sectors according to the closest available TPSO price categories. For the Paper sector, the matched Paper and Paper Products price index was 104.3, indicating that the corresponding 2021 monetary values were deflated by a factor of 1.043 . For the Rubber, Chemical sector, which combines petroleum refining, chemical, and rubber and plastic activities, an index of 104.5 was applied based on the available matched price categories. For sectors without a directly corresponding TPSO price category, including PubU, Construction, Transportation, and T\&S, the common All (Proxy) index of 102.3 was applied. The Unclassified sector was assigned an index of 100.0, implying no price adjustment relative to the 2015 reference basis. The resulting sector-matched deflators were consistently applied to the intermediate transaction matrix, total output, and final demand vector to express the 2021 IO data in 2015 constant prices before conducting the temporal comparison and structural decomposition analysis (SDA) [19-21].

Table 2. Sector-matched price deflators used to convert the 2021 input–output (IO) data to 2015 constant prices

Note: TPSO = Trade Policy and Strategy Office.

2.5 Estimation of greenhouse gas emissions from fossil fuel combustion

Direct GHG emissions from fossil fuel combustion are estimated for each sector using fuel consumption data and corresponding emission factors. Fuel consumption data were converted from the reporting units used in the national energy statistics to the units required by the corresponding emission factors using the fuel-specific conversion factors documented in the calculation files. Fuel consumption, denoted as $F_{i, f}$, represents the amount of fuel type $f$ consumed by sector $i$. The emission factor, denoted as $E F_f$, represents the amount of emissions generated per unit of fuel consumed. The resulting emissions are expressed in kg $\mathrm{CO}_2$ and represent sectoral direct emissions associated with fossil fuel combustion. The use of fuel-specific emission factors enables differentiation across fuel types with varying carbon intensities, thereby improving the accuracy and consistency of emission estimation in sectoral analysis [14]. This formulation is widely applied in energy and environmental studies to quantify emissions at the sector level prior to integration into IO models. The direct emissions are calculated as:

$E_i=\sum_f\left(F_{i, f} \times E F_f\right)$   (3)

where, $E_i$ represents total direct GHG emissions from sector $i$, aggregated across all fuel types $f$. These emissions represent direct fossil-fuel combustion emissions occurring within each sector and are distinguished from final-demand-based embodied emissions calculated subsequently through the IO framework.

2.6 Construction of emission coefficients

Emission coefficients are calculated to integrate environmental information into the IO framework. For each sector $i$, the emission coefficient is defined as the ratio of direct GHG emissions to total sectoral output, as expressed in Eq. (4). Specifically, $e_i$ represents the emission coefficient, $E_i$ denotes direct GHG emissions from fossil fuel combustion in sector $i$, and $x_i$ is the total economic output of that sector.

$e_i=\frac{E_i}{x_i}$   (4)

The resulting coefficient $e_i$ represents the amount of emissions generated per unit of economic output (kg CO_₂e/THB). These coefficients represent emissions per unit of economic output and are commonly used in environmentally extended IO models [7, 9]. They enable the integration of environmental data with economic systems and facilitate embodied emission analysis [11, 18]. The emission coefficient therefore represents the direct fossil-fuel combustion emissions per unit of sectoral economic output and serves as the environmental extension coefficient for calculating final-demand-based embodied emissions in the IO framework.

For 2021, the emission coefficient was recalculated using constant-price sectoral output. Physical direct GHG emissions $E_i$, derived from fuel consumption and emission factors, were not deflated because they represent physical emission quantities.

2.7 Input–output-based embodied emissions

Total GHG emissions associated with economic activities are estimated by integrating emission coefficients with the IO framework:

$C=e(I-A)^{-1} y=e L y$   (5)

where, $C$ represents total GHG emissions, $e$ is the vector of emission coefficients, and $L=$ $(I-A)^{-1}$ is the Leontief inverse matrix, and $y$ is the final-demand vector. The formulation captures the direct and indirect GHG emissions required to satisfy final demand through the modeled domestic production system [7, 22]. The resulting emissions are therefore interpreted as final-demand-based embodied emissions rather than direct production emissions, which are calculated separately from sectoral fossil fuel consumption in Section 2.5.

For the economy-wide assessment, $y$ contains the final-demand values of all 14 sectors, and $C=e L y$ represents the total GHG emissions embodied in aggregate final demand. For a specific sector $j$, sector-specific embodied emissions are obtained by constructing a sector-specific final-demand vector $y_j$, in which the final-demand value of sector $j$ is retained and the final-demand values of all other sectors are set to zero. The corresponding sector-specific embodied emissions are calculated as

$C_j=e L y_j$   (6)

where, $C_j$ represents the GHG emissions embodied in the final demand of sector $j$.

For paper products, the paper-specific final-demand vector is defined as

$y_{\text {Pap }}=s_{\text {Pap}} y_{\text {Pap}}$   (7)

where, $s_{\text {Pap}}$ is a $14 \times 1$ sector-selection vector with a value of 1 for the Paper (Pap) sector and 0 for all other sectors, and $y_{\text {Pap}}$ is the final-demand value of the Paper sector. Thus, $y_{\text {Pap}}$ contains the Paper-sector final demand while all other sectoral final-demand components are set to zero. The paper-specific embodied GHG emissions are subsequently calculated as

$C_{\text {Pap}}=e L y_{\text {Pap}}$   (8)

This result represents the direct and indirect GHG emissions embodied in the domestic production requirements associated with paper-sector final demand. Because the model is based on a domestic single-region IO table, imported intermediate inputs are not traced to their foreign producing sectors, and their upstream fossil-fuel emissions are therefore not included in the estimated embodied emissions. The paper-specific embodied-emission result should be distinguished from the final-demand effect obtained from the SDA. Specifically, $C_{\text {Pap}}$ represents the total embodied emissions associated with paper-sector final demand, whereas $\Delta C_y$ represents the economy-wide contribution of changes in final demand to the change in total GHG emissions between 2015 and 2021.

For the 2021 constant-price framework, the total economy-wide emissions were calculated using the price-harmonized variables as $C_{2021}=e_{2021}^{C P} L_{2021}^{C P} y_{2021}^{C P}$, where the emission coefficient, Leontief inverse, and final-demand vector were consistently derived from the price-harmonized IO framework.

2.8 Emission intensity analysis

Emission intensity is defined as the amount of GHG emissions generated per unit of production output and is used to assess the carbon efficiency of production systems. For paper products, emission intensity is calculated as the ratio of paper-specific embodied GHG emissions, obtained from the paper final-demand vector, to the corresponding paper-product production volume, as expressed in Eq. (9). The production volume represents the total output of paper products and excludes pulp production.

$E I_{\text {Pap}}=\frac{C_{\text {Pap}}}{Q_{\text {Pap}}}$   (9)

Specifically, $E I_{\text {Pap}}$ represents the emission intensity of paper products, $C_{\text {Pap}}$ denotes the GHG emissions embodied in the final demand of the Paper sector and $Q_{\text {Pap}}$ denotes the corresponding paper-product production volume.

The resulting emission intensity is expressed in units of t CO₂e per tonne of product (t CO₂e/t) and reflects the overall carbon efficiency of the production system. Emission intensity is a key indicator for evaluating carbon efficiency and is widely used in environmental performance assessment [24, 25]. It enables benchmarking across sectors and comparison with national standards [3, 14].

2.9 Temporal comparison (2015 and 2021) and structural decomposition analysis

A comparative analysis is conducted to examine changes in emissions between 2015 and 2021. This approach is widely used in SDA to identify key drivers of emission changes [8, 26]. By maintaining a consistent framework, the analysis enables robust evaluation of demand, structural, and technological effects [7, 26].

A comparative analysis was conducted to examine changes in GHG emissions between 2015 and 2021 using a common price basis. The 2015 IO data were retained at the reference price basis, while the 2021 monetary IO data were harmonized to 2015 constant prices using sector-matched price deflators.

The temporal comparison between 2015 and 2021 was conducted using a common 2015 constant-price basis. The 2015 IO table was retained as the reference price structure, while the 2021 monetary IO data were harmonized using sector-matched producer or service price deflators derived from the closest available TPSO price indices. The price-harmonized 2021 intermediate transactions, total output, and final demand were then used to recalculate the technical coefficient matrix and the Leontief inverse. Accordingly, monetary changes in final demand between the two years are interpreted on a constant-price basis rather than as changes in current-price values [20, 21].

To identify the determinants of the change in total GHG emissions, an additive SDA was applied to the IO identity $C=e L y$. Let superscript 0 denote 2015 and superscript 1 denote 2021 after price harmonization. The decomposition separates the observed change in total emissions into emission-coefficient, Leontief-inverse (production-structure), and final-demand effects. Because SDA can yield multiple equivalent decompositions depending on the ordering of factors, the present study uses the average of the two polar decompositions, a conventional approach that reduces dependence on a single factor-ordering scheme [11, 26, 27].

$C^t=e^t L^t y^t$   (10)

$\Delta C=C^1-C^0=e^1 L^1 y^1-e^0 L^0 y^0$   (11)

$\Delta C^e=\frac{1}{2} \Delta e\left(L^1 y^1+L^0 y^0\right)$   (12)

$\Delta C^L=\frac{1}{2}\left[e^1 \Delta L y^0+e^0 \Delta L y^1\right]$   (13)

$\Delta C^y=\frac{1}{2}\left[e^1 L^1 \Delta y+e^0 L^0 \Delta y\right]$   (14)

$\Delta C^e+\Delta C^L+\Delta C^y=\Delta C$   (15)

$\Delta e=e^1-e^0, \Delta L=L^1-L^0, \Delta y=y^1-y^0$   (16)

2.10 Uncertainty and limitations

The IO-based emission intensity is compared with national benchmark values derived from the Carbon Footprint of Products (CFP) database developed by TGO. The comparison is used solely as a contextual benchmark for interpreting the magnitude of the IO-based estimate rather than as a formal validation of the IO model.

Differences between the IO-based estimates and the CFP benchmark may arise from differences in temporal coverage, product composition, allocation procedures, system boundaries, and sectoral aggregation. The IO approach represents a top-down economic framework that captures emissions associated with intersectoral production linkages, whereas the CFP benchmark represents product-level emission factors developed under the TGO CFP assessment framework. Therefore, the two estimates are not considered directly equivalent measures, and their quantitative difference is not interpreted as evidence of model validity or invalidity.

The analysis is limited to GHG emissions from fossil fuel combustion and excludes other emission sources such as industrial process emissions, agriculture, land-use change, and waste management, consistent with IPCC guidelines [16, 17]. This restriction ensures methodological consistency but may lead to partial representation of total sectoral emissions.

Additional uncertainties arise from the use of IO tables, sectoral aggregation, and estimated production data. These uncertainties are inherent in IO-based environmental assessments and have been widely discussed in the literature, particularly in the context of multi-region input–output (MRIO) models applied to national carbon footprint analysis [11, 28].

While such uncertainties may influence the magnitude of estimated results, previous studies demonstrate that IO-based approaches remain robust for macro-level analysis and comparative assessments, especially when consistent datasets and methodological frameworks are applied [11, 28].

This study confirms that the IO framework is a robust and policy-relevant tool for evaluating GHG emission intensity at the national level, particularly for paper products, which are characterized by complex intersectoral linkages and high energy dependence. By integrating national energy statistics with IO tables, the approach captures both direct and indirect emissions associated with domestic intersectoral production requirements, enabling a comprehensive understanding of emission drivers within the domestic production system.

The results indicate that the emission intensity of paper products increased between 2015 and 2021. The temporal comparison was conducted using monetary IO data harmonized to a common 2015 price basis, allowing changes in monetary final demand to be interpreted at constant prices. The SDA demonstrates that the observed economy-wide increase in GHG emissions resulted from substantial and offsetting contributions from final demand, production structure, and emission coefficients. Specifically, the final-demand effect contributed +3.375 Mt CO₂e, the Leontief-inverse / production-structure effect contributed −14.405 Mt CO₂e, and the emission-coefficient effect contributed +14.468 Mt CO₂e, yielding the observed net increase of 3.438 Mt CO₂e. These results indicate that the net emission increase cannot be attributed to final demand alone and that mitigation strategies should address demand-side changes, intersectoral production requirements, and fossil-fuel emission coefficients jointly.

The IO-based emission intensities differ from the weighted CFP benchmark, reflecting differences in system boundaries, temporal coverage, product composition, allocation procedures, and sectoral aggregation. Therefore, the CFP comparison is presented solely as a contextual benchmark rather than as a formal validation of the IO-based estimates. The two approaches are complementary: the IO framework provides an economy-wide perspective on direct and indirect emissions associated with intersectoral production linkages, whereas the CFP framework provides product-level emission information.

However, several limitations should be noted, including the exclusion of non-fossil emission sources, the use of sector-matched price deflators where directly corresponding TPSO indices were unavailable, the estimation of production data for 2015 due to data constraints, and the use of a domestic single-region IO framework. Consequently, foreign upstream GHG emissions embodied in imported intermediate inputs are not explicitly captured. The results should therefore be interpreted as emissions associated with domestic production requirements in Thailand rather than as a complete global supply-chain or consumption-based carbon footprint. These limitations may introduce uncertainty into the estimates and should be considered when interpreting the results. Future research should focus on improving data harmonization, incorporating more recent datasets, and developing multi-region or hybrid input–output and Life Cycle Assessment (IO-LCA) approaches to enhance coverage and resolution. Overall, the IO framework provides a complementary economy-wide perspective on the structural determinants of GHG emissions associated with paper products and supports evidence-based policy development for sustainable industrial transformation in Thailand.