Exploring CO₂ emission factors in Vietnam towards sustainable consumption

Sustainable development (SD) is becoming more important because of global climate challenges (Prizzia, 2017). Higher temperatures and harsh weather have made more people aware of how SD can help lessen environmental harm.

Sustainable development must tackle environmental and climatic issues (Shah et al., 2024), and cutting CO₂ emissions is a crucial solution. Given this, Vietnam signed the Paris Agreement and joined in the worldwide battle against climate change. During the 26th Conference of the Parties (COP26), Vietnam pledged to take action against climate change. However, rapid economic growth has caused Vietnam’s CO₂ emissions to exceed the global average. Research by Thu et al. (2022) indicates that a 1% increase in GDP corresponds to a 1.26% increase in emissions in Vietnam. Baltagi et al. (2019) found a complex relationship between CO₂ emissions and GDP, with the recovery rate of CO₂ emissions nearly equal to GDP growth. Such findings highlight significant challenges to sustainable development in Vietnam.

Carbon emissions can be quantified using either production-based or consumption-based approaches (Peters, 2008; Davis and Caldeira, 2010). In life cycle analysis, emissions and resource use during production relate to the final consumption of goods and services. Consumption-based emissions (CBE) calculations provide a valuable complement to production-based emissions (PBE) calculations by accurately tracking emissions along the global production chain driven by consumer demand (Peters, 2008; Peters et al., 2011). Therefore, to avert adverse environmental impacts inhibiting the ability of the economy and human society to grow sustainably, every country needs to find ways to quantify and decrease CO₂ emissions from consumption patterns.

Compared with other nations, final consumption activities significantly contribute to pollution, especially CO₂ emissions, in Vietnam. According to Nguyen et al. (2018), the structure and extent of consumption exacerbate CO₂ accumulation, with investment and household activities as the primary sources. The main sources of emissions are energy use in the residential and manufacturing sectors, where electricity consumption is a crucial determinant (Nguyen et al., 2021; Raihan, 2023). Forecasts indicate that Vietnam’s urban population share will exceed 50% by 2030 (General Statistics Office, 2023). In parallel, average household income grew by about 7% per year between 2010 and 2020 (General Statistics Office, 2021), signalling a surge in electricity, cement and private-vehicle demand if left unchecked. Coupled with Vietnam’s Net-Zero-2050 pledge and the forthcoming EU Carbon Border Adjustment Mechanism (European Commission, 2023), reducing domestic and export-embedded carbon footprints has become essential for sustaining trade competitiveness. Taken together, these factors make Vietnam an ideal case study.

Our research contributes by providing new insights for the literature, managers, and policymakers in the following aspects: (1) It is the first comprehensive analysis in Vietnam using SDA and SAM models with the latest data to examine the effect of factors on CO₂ emissions from final consumption. This combination approach informs policies focused on reducing emissions and addressing income distribution; (2) The study evaluates the effect of factors on CO₂ emissions in different types of final consumption, such as household, government, investment, inventory, and exports, together rather than assessing each type separately, as in previous studies. Such an analysis provides a comprehensive view of the relationship between final consumption and emissions in the economy; (3) It serves as a case study for reducing carbon emissions from final consumption in developing countries, offering a reference for nations with similar contexts to Vietnam to achieve sustainable consumption, mitigate environmental damage, and foster economic growth. Thus, this study seeks to answer the following research question: “How have changes in final consumption (household, government, investment, inventory and exports), along with variations in emission intensity and economic structure, contributed to changes in CO₂ emissions in Vietnam over the period 2007–2020?”.

The remainder of the research is structured as follows: Section 2 reviews the relevant literature. Section 3 outlines the data and methodology employed in this research. Section 4 presents the results and discussion, while Section 5 highlights the key conclusions with recommendations for relevant policy actions.

“Sustainable development,” as defined by the United Nations, is understood as “development that meets the needs of the present, without compromising the ability of future generations to meet their own needs” (World Commission on Environment and Development, 1987). Accordingly, development must maintain a balanced and harmonious integration of all three pillars: economic, social, and ecological/resource-environmental development. To establish SD, the integration of consumption and production systems with SD has been designed and implemented (Lozano et al., 2015). In 1994, the Oslo Symposium on Sustainable Consumption presented the first formal definition of Sustainable Consumption and Production (SCP). The definition states that achieving SCP requires consuming in a way that meets people’s basic needs, enhances the quality of life, and simultaneously minimizes environmental impact by reducing the use of natural resources, harmful materials, waste, and pollutant emissions throughout the entire lifecycle of products and services. The goal is to ensure that the needs of future generations are not compromised (Ofstad et al., 1994). By 2015, SCP had become one of the most critical components of the 2030 Agenda for Sustainable Development. From a consumption perspective, SCP implies changing consumption patterns of households and governments through shifts in lifestyle, behaviour, and individual consumer choices, as well as changes in public sector procurement strategies (World Bank, 2017). Sustainable consumption involves essential solutions for transforming production and consumption patterns (Lukman et al., 2016). Consumers will shape demand for environmentally responsible products and services by choosing more sustainable options. As a result, businesses will be encouraged to adopt more sustainable practices.

The final consumption substantially contributes to CO₂ emissions. Household consumption contributes 40–70% in developed countries (Wang et al., 2024) and 48–85% in developing economies (Shershunovich and Mirzabaev, 2024). In developing countries, food and housing dominate greenhouse gas emissions, while in industrialised nations, housing, mobility, food, and electrical appliances account for over 70% of household consumption’s impact (Hertwich et al., 2010). While household consumption is a direct driver of CO₂ emissions, it also contributes to broader environmental impacts such as oversized ecological footprints, particularly through high levels of energy use. Using a fuzzy-set Qualitative Comparative Analysis for 103 countries, Kazemzadeh et al. (2023) show that oversized ecological footprints emerge from specific configurations that couple high GDP per capita, rapid urbanisation and large energy consumption, underscoring the need for frameworks able to track multiple drivers simultaneously. Smaller households may have higher per capita consumption, leading to increased emissions (Zhang et al., 2023; Shao et al., 2024). Higher-income households typically lead to higher carbon emissions due to greater consumption of carbon-intensive goods (Yologlu and Halisdemir, 2024). However, improving energy efficiency is one approach to mitigate this impact (Sun et al., 2024). Evidence for EU-14 likewise indicates that wider gender pay gaps, together with higher energy use, raise CO₂ emissions, confirming the amplifying role of social inequality in household carbon footprints (Koengkan and Fuinhas, 2021). Further panel evidence by Kazemzadeh et al. (2022) shows that faster progress in the energy transition consistently lowers per-capita CO₂. However, “brain drain” can partly erode these gains, emphasising the joint role of technology and human capital.

Government consumption accounts for approximately 10% of global greenhouse gas emissions (Hertwich and Peters, 2009), and in the UK, this figure is nearly 14% (Wiedmann et al., 2007). In Europe, these emissions primarily stem from administrative activities, defence, education, healthcare, and elderly care (Moll et al., 2007). Asian economies are experiencing increasing emissions due to substantial investments in infrastructure. However, government consumption on green initiatives and healthcare can significantly reduce emissions (Taher, 2024). Furthermore, the distribution of expenditures toward education, science, R&D, and green finance can substantially enhance environmental efficiency (Bilgili et al., 2021; Jin et al., 2022).

A considerable body of research explores the complex interaction between trade activities and CO₂ emissions, identifying positive and negative effects. Fang et al. (2019) demonstrate that for developing economies, expanded export activities and fewer trade restrictions foster economic growth and contribute to an increase in per capita CO₂ emissions. Developed countries mitigate export-related emissions through advanced technologies while importing diversification reduces emissions in developed nations but increases them in emerging markets reliant on non-renewable energy (Hu et al., 2020). Global trade aggravates the carbon footprints of the building-and-construction segment, thus fostering efforts against carbon leakage and green supply chain introduction (Gao et al., 2022).

The significance of the relationship between inventory management and CO₂ emissions has received little attention. The more stockpile businesses carry, the more transportation and storing activities they engage in, so the more emissions they generate (Caballero-Morales and Martínez-Flores, 2020; Ni et al., 2024). Mashud et al. (2021) found that implementing preservation technologies reduced emissions from inventory control.

Countries with low GDP per capita, where cleaner technologies are less available, tend to emit higher CO₂ levels, with FDI having an adverse correlation with emissions. This supports the pollution haven hypothesis (Xie et al., 2020; Abdullahi et al., 2023; Wang et al., 2023). On the other hand, developed countries use FDI to invest in cleaner technology, contributing to reducing pollution. FDI has a wide range of impacts on different regions, so supportive policies need to be established to achieve sustainability objectives (Mesagan et al., 2021).

SDA is a quantitative tool that dissects the change in CO₂ emissions into factors like changes in technology, changes in consumption, and institutional structure configuration (Miller and Blair, 2009). The SDA approach, together with the Input-Output (IO) one, goes beyond the intersectoral analysis in the sense that they provide information on the determinants of CO₂ emissions from energy use and sectoral demand (Hoekstra and van den Bergh, 2003; Wang et al., 2014; Cansino et al., 2016; Lin et al., 2022; Liu et al., 2023b). In this respect, Sun and Mi (2023) examined the impact of COVID-19 on carbon emissions changes, stating that the changes in production processes and larger shares of carbon embedded in materials increased carbon emissions and stressed the need for more investment in low-carbon solutions, along with a transition towards consumption-led economic recovery. Expanding this approach to the Social Accounting Matrix (SAM) improves understanding of the relationship between emissions and social and production agents. The SAM framework reveals significant aspects of household structure and household consumption behaviour in the economy. Moreover, the results of research combining SDA and SAM by Dirk van Seventer and Finn Tarp (2023) and Álvarez-Martínez and Mainar-Causapé (2021) improve SDA and IO results, reducing mistakes and improving the policy process. However, only a limited number of studies integrate SDA with SAM.

The database used in this research consists of the Social Accounting Matrix (SAM) tables for Vietnam for 2007, 2012, 2016, and 2020. The years 2007, 2012, 2016, and 2020 were selected as they represent key turning points in Vietnam’s development: 2007 marks post-WTO industrialisation, 2012–2016 reflects heavy-industry expansion, and 2020 captures the COVID-19 shock. These years also align with the availability of OECD 45-sector IO data up to 2020, enabling a consistent analysis of CO₂ emission trends. The SAM tables were updated and constructed following the CIEM approach (a method promoted by the Central Institute for Economic Management) and (Nguyen et al., 2023), based on the following data sources: the country’s Input-Output (IO) data from the OECD; National Accounts data from the General Statistics Office; State Budget revenue and expenditure settlements from the Ministry of Finance; Balance of Payments data from the State Bank; and the Vietnam Household Living Standards Survey. The RAS method was employed to carry out the balance of accounts in the micro SAM.

Vietnam’s economic sectors consist of 45 industries according to the OECD sector classification, including: “C1 – Agriculture, hunting, forestry; C2 – Fishing and aquaculture; C3 – Mining and quarrying, energy producing products; C4 – Mining and quarrying, non-energy producing products; C5 – Mining support service activities; C6 – Food products, beverages and tobacco; C7 – Textiles, textile products, leather and footwear; C8 – Wood and products of wood and cork; C9 – Paper products and printing; C10 – Coke and refined petroleum products; C11 – Chemical and chemical products; C12 – Pharmaceuticals, medicinal chemical and botanical products; C13 – Rubber and plastic products; C14 – Other non-metallic mineral products; C15 – Basic metals; C16 – Fabricated metal products; C17 – Computer, electronic and optical equipment; C18 – Electrical equipment; C19 – Machinery and equipment, nec; C20 – Motor vehicles, trailers, and semi-trailers; C21 – Other transport equipment; C22 – Manufacturing nec; repair and installation of machinery and equipment; C23 – Electricity, gas, steam, and air conditioning supply; C24 – Water supply; sewerage, waste management and remediation activities; C25 – Construction; C26 – Wholesale and retail trade; repair of motor vehicles; C27 – Land transport and transport via pipelines; C28 – Water transport; C29 – Air transport; C30 – Warehousing and support activities for transportation; C31 – Postal and courier activities; C32 – Accommodation and food service activities; C33 – Publishing, audiovisual, and broadcasting activities; C34 – Telecommunications; C35 – IT and other information services; C36 – Financial and insurance activities; C37 – Real estate activities; C38 – Professional, scientific, and technical activities; C39 – Administrative and support services; C40 – Public administration and defense; compulsory social security; C41 – Education; C42 – Human health and social work activities; C43 – Arts, entertainment and recreation; C44 – Other service activities, C45 – Activities of households as employers; undifferentiated goods- and services-producing activities of households for own use”.

The household data in Vietnam’s micro-SAM comprises ten groups in both urban and rural areas, representing household income levels increasing from H1 to H5 in urban and from H6 to H10 in rural. Every group represents 20% of the households in each area.

In plain terms, ESAM starts from the usual Social Accounting Matrix (SAM) of economic transactions and “attaches” an emissions-intensity matrix (⁠$B$⁠), where each entry tells how many tonnes of CO₂ are released per unit of output for each sector. By multiplying B by the SAM multiplier matrix (⁠$Ma$⁠) and the final-demand vector (⁠$y$⁠), we calculate total CO₂ emissions for each type of final consumption. Structural Decomposition Analysis (SDA) then breaks down changes in these emissions into three key drivers—final-demand scale (⁠$∆Fy$⁠), emission intensity (⁠$∆FB$⁠) and economic-structure shifts (⁠$∆FMa$⁠). Figure 1 illustrates this step-by-step process.

Our ESAM extends the classic input–output framework (Miller and Blair, 2009), thereby combining economic transactions with environmental emissions in a single model. This approach therefore estimates the links between pollutants and economic activities more precisely. Furthermore, it indicates that changes in income and consumption levels lead to changes in CO₂ emissions, stating that the linear correlations between the endogenous and exogenous accounts in SAM can be derived as represented in the equations below:

In the context of this problem, $X$ is the vector of total output from production activities; $I$ is the identity matrix; and $A$ is the matrix of average endogenous expenditure propensities, where the components of the matrix $A$ are derived by dividing the transactions in the SAM by the total column of the accounts; $y$ is the final demand matrix, encompassing household and government consumption, investment, inventory, and exports.

We apply the ESAM–SDA approach to quantify CO₂ emissions by final-demand category, treating domestic output as endogenous and households and factors as exogenous, and decompose changes into each group’s contribution. In this way, the role of household consumption in CO₂ emissions will become more evident, along with the influence of remaining demand groups. Accordingly, the carbon emissions embedded in the final consumption (⁠$y$⁠) in the SAM model can be described by the following equation:

Here, $F$ is the vector of emissions satisfying the final consumption $y$⁠, and $B$ is the vector determined by CO₂ intensity, representing the emissions per unit of output for each sector. The SAM multiplier matrix (⁠$Ma$⁠) quantifies the total impact on endogenous income related to a one-unit change in exogenous impact.

For this investigation, SDA was used to analyse such changes in total carbon emissions to component carbon changes generated from emission intensity (⁠$dFB$⁠), IO structure (⁠$dFMa$⁠), and final consumption (⁠$dFy$⁠). Therefore, the Eq. (2) can be decomposed as shown below:

In particular, the impact of emission change derived from these factors is expressed anew in Eq. (3) which involves the decomposition of Eq. (2) in the following way:

The notation $∆F$ applies to the total change in CO₂ emissions. $∆FB$⁠, $∆FMa$ and $∆Fy$ are the carbon emission changes affected by changes in emission intensity (⁠$∆B$⁠), IO structure (⁠$∆FMa$⁠), and final consumption (⁠$∆Fy$⁠), respectively.

The “polar decomposition method” (Jinghua, 2005) is applied to reveal the factors influencing F. Accordingly, $∆FB,∆FMa,∆Fy$ can be represented by the equations from (5) to (7):

Where $∆B=B1−B0,∆Ma=Ma1−Ma0,∆y=y1−y0$

The subscripted 0 and 1 indicate the initial and the final points in time.

Vietnam stands out as one of the fastest-growing economies in Southeast Asia. According to Liu et al. (2023a), Vietnam contributed the second-highest total CO₂ emissions in Mainland Southeast Asia in 2017, following Thailand, with both countries accounting for over 80% of the region’s emissions. Vietnam also exhibited a rapid average annual emission growth rate of 17.9% during 2010–2019, second only to Laos (19.3%), and far above Thailand (2.5%). This combination of high volume and fast growth highlights Vietnam’s unique emission profile in the region and reinforces the urgency for structural mitigation policies.

GDP grew approximately 13% annually from 2007 to 2020, driving higher consumption and production and thus raising CO₂ emissions (Table 1). Exports reflecting Vietnam’s deeper global integration accounted for the largest share of final consumption and CO₂ emissions. Household consumption and investment were the second and third-largest contributors, respectively. Although resource use remained relatively stable, CO₂ emissions from activities utilizing those resources remained high. Government consumption had a negligible share, reflecting tight fiscal control, while only inventory showed significant emission reductions in 2020 due to pandemic-driven demand drops.

In analysing final consumption, we group sectors into: (1) those consistently among the top 10 contributors to CO₂ emission changes across all three periods (2007–2012, 2012–2016, 2016–2020); and (2) those appearing in the top 10 only once or twice. The former reflects persistent drivers needing long-term policy action, while the latter captures short-term spikes requiring timely and targeted responses.

Final consumption (⁠$Fy$⁠) was the primary driver of total emissions growth across all three periods. From 2007 to 2016, $Fy$ consistently rose due to expanding domestic consumption and export, especially the rise of inventory in 2012–2016 due to economic recovery after the crisis. However, by 2016–2020, the growth in Fy slowed significantly compared to previous periods, indicating stabilization in consumption growth rates, primarily due to the global economic downturn triggered by COVID-19, which altered domestic consumption patterns.

While emission intensity (⁠$FB$⁠) was negative in the first two periods, its absolute value declined, indicating that emission intensity became less effective in reducing CO₂ emissions over time. By 2016–2020, $FB$ turned positive, marking a reversal as emission intensity began to contribute to emissions growth. This shift reflects the continued reliance on high-emission industries such as steel and coal power, limited adoption of cleaner technologies, and COVID-19-related disruptions that delayed energy-efficiency upgrades.

The economic structure (⁠$FMa$⁠) was strongly negative in 2007–2012 as CO₂ emissions from

C22 and C23 fell sharply due to process optimization, automation, and energy-efficiency policies, alongside a shift toward higher-value, lower-emission services. It turned positive in 2012–2016 when emissions from C14 and C28 surged under a wave of BOT/PPP infrastructure projects and FDI incentives, with traditional kiln and port technologies driving higher fuel and material use. From 2016 to 2020, $FMa$ reverted to negative as Green Growth Strategy 2021–2030 and national digitalization encouraged energy-efficient kilns and port automation, while COVID-19 restrictions in early 2020 forced process streamlining, together cutting emissions in these sectors.

Household consumption and exports were the most volatile drivers of CO₂ emissions, reflecting their sensitivity to economic cycles and structural change. Investment spikes in 2007–2012 and again in 2016–2020, driven by capital-intensive infrastructure projects and COVID-19 stimulus, also boosted emissions, whereas government consumption remained broadly unchanged. Inventory movements showed the sharpest reversal: restocking in 2012–2016 raised emissions, but a −117,843 billion tCO₂ drop in 2016–2020, as firms drew on existing stockpiles amid pandemic disruptions, made inventory the sole final-demand category to cut emissions in that period.

The sectors C23 and C15 are among those with significant discrepancies in CO₂ emissions arising from household consumption activities (Table 2). Emissions from C23 are highest, indicating a strong demand for energy within households, particularly in electricity and air conditioning usage. Sector C15 also exhibits substantial emissions, although lower than those from C23, reflecting the growing demand for infrastructure development amid rising urban growth. The emission intensity (⁠$FB$⁠) of these two sectors also contributes to the overall rise in emissions. Although structural adjustments (⁠$FMa$⁠) within the economy have helped reduce emissions from these sectors at certain stages, the reduction has been negligible compared to the overall increase in emissions from both sectors (Table 3).

The CO₂ emissions from the two sectors, C23 and C15, also display significant fluctuations across different household groups and regions in all three periods (Table 3). Households with low income (H1, H2, H6, and H7) generally emit less compared to those with middle (H3 and H8) and high income (H4, H5, H9, and H10) within the same region, primarily due to their limited consumption capacity because of being forced to focus on products of essential needs. The H5 group recorded emissions from sector C23 significantly declined between 2012 and 2016, thanks to the trend of applying renewable energy sources such as solar energy and the effective implementation of energy-saving measures. However, emissions began to rise again due to heightened consumption demands during the COVID-19 lockdown. In rural areas, despite lower incomes compared to urban, households tend to generate higher total emissions than urban households due to their larger populations, resulting in higher energy consumption demands.

Outside C23 and C15, household consumption drove CO₂ emissions across other sectors, depending on consumption demand of each period, such as C2, C6, C7, C11, C14, C27, C29, and C32 from 2007–2012; C1, C4, C5, C10, C20, C25, C28, and C41 from 2012–2016; and C1, C6, C9,C10, C11, C14, C22, and C26 from 2016–2020. This trend reflects the increased demand of households in construction, food, urban lifestyles, transportation, and essential services.

From 2007 to 2020, government consumption significantly impacted CO₂ emissions in sectors C14 and C15 (Table 2). The increase in emissions from these two economic sectors primarily stemmed from substantial public spending (⁠$Fy$⁠) on infrastructure, along with natural resource extraction and energy intensity (⁠$FB$⁠); however, the economic structure (⁠$FMa$⁠) was insufficient to mitigate environmental impacts (Table 3).

In these periods, several economic sectors also recorded significant increases in CO₂ emissions due to government consumption, such as C3, C9, C11, C12, C29, C32, C34, and C42 in 2007–2012; C1, C4, C10, C23, C26, C27, C28, and C29 in 2012–2016; and C4, C9, C10, C11, C12, C22, C23, and C40 in 2016–2020. These findings indicate that despite policies aimed at promoting renewable energy and sustainable service sectors, CO₂ emissions did not decrease due to the expansion of the government’s concentrated spending on transportation infrastructure, traditional industries, and heavy industries.

Investments from 2007 to 2020 have significantly contributed to emissions in sectors C4, C10, C15, C23, and C25 (Table 2). Heightened investment demand (⁠$Fy$⁠) in energy, construction, construction materials, and natural resource extraction fulfils export and domestic needs but contributes to emissions growth. The rise in $FB$ of all sectors also drives CO₂ emissions. Only for C25, $FB$ decreased in 2012–2016 thanks to applying energy-saving and emission-reduction policies and green construction technologies. Despite that, large-scale deployment of green construction technologies remains challenging due to high costs and limitations, leading to higher CO₂ emissions in the following period (Table 3).

Investment also significantly influenced CO₂ emissions in various sectors depending on each period, including C3, C11, C14, C27, and C29 in 2007–2012; C1, C21, C27, C28, and C29 in 2012–2016; C9, C11, C14, C22, and C26 in 2016–2020. This data shows that investments mainly went into the infrastructure of industries that consume large amounts of energy, particularly those related to processing and trade.

Significant increases in CO₂ emissions resulted from the effects of inventory on sectors C22, C26, and C27 between 2007 and 2020 (Table 2). Improved inventory control (⁠$Fy$⁠) and modifications in emission intensity (⁠$FB$⁠) are the primary causes that induced a steady decline in CO₂ emissions for C22 and a sharp decrease in CO₂ emissions for C26 and C27 during 2007–2012 (Table 3). However, from 2012 to 2020, C26 and C27 experienced a considerable increase in emissions, driven by the intensified investment and large-scale production, which led to a resurgence in inventory levels.

Other sectors such as C31, C34, C36, C37, C38, C40, C41, C43, and C44 from 2007–2012; C3, C7, C10, C11, C14, C23, C26, C27, and C29 from 2012–2016; and C1, C2, C8, C25, C28, C32, C39, and C44 from 2016–2020 recorded significant emission increases from inventory, associated with expanding production and investment in large-scale projects, particularly intensive industries, and the need to stockpile raw materials, leading to increased inventory requirements and energy consumption.

The export trend has resulted in a notable rise in emissions in sectors C15, C23, C27, and C29 across the three analysed periods (Table 2). CO₂ emissions from exports of these sectors were primarily influenced by increasing export demand (⁠$Fy$⁠), while emission intensity (⁠$FB$⁠) and economic structure (⁠$FMa$⁠) exhibited different fluctuations. In some periods, the reduction in FB for sectors C23, C27 and C29 indicates the effective implementation of new technologies or more efficient production processes, resulting in lower emissions (Table 3).

Furthermore, all three periods also witnessed significant increases in emissions from several other industries, such as C3, C6, C7, C9, C11, and C14 during 2007–2012; C1, C4, C5, C10, C25, and C28 during 2012–2016; and C4, C7, C10, C11, C14, and C28 during 2016–2020. These findings show that Vietnam’s export-driven industrial expansion has spurred emissions, while clean-technology adoption remains slow to meet its sustainability goals.

To validate the robustness of our SDA results at the sector–consumption category level, we compared the Midpoint, IDA, and Shapley decompositions (Miller and Blair, 2009) against the Polar baseline (Tables IV to X in supplementary file). Midpoint—by averaging f, L, and y at the cycle midpoint—smooths nonlinear fluctuations and yields a decomposition direction almost identical to Polar. In contrast, IDA relies exclusively on base-year values (f₀, L₀, y₀), sometimes failing to capture structural shifts and producing deviations in both magnitude and direction. Shapley’s permutation approach is highly sensitive to interactions and nonlinearities, resulting in the greatest number and magnitude of mismatches.

The results show that only Midpoint fully aligns with the Polar trend; 8 of 45 sectors under IDA and 33 of 45 under Shapley exhibited directional mismatches relative to Polar. In three 2019 interpolation scenarios (75% 2016 + 25% 2020 to mitigate the COVID-19 shock; 50–50% midpoint; and 25% 2016 + 75% 2020 to amplify pandemic effects), only Shapley recorded mismatches in 23 of 45 sectors, consistently across all weights. These findings pinpoint which sectors and end-use categories are method-sensitive, without undermining our central conclusion that final demand remains the dominant driver or the overall stability of the SDA outcomes.

The research examines the impact of final consumption on CO₂ emissions in Vietnam, leading to the following key findings:

1. Exports are the largest contributor, reflecting Vietnam’s integration into global value chains. While emissions from inventory have declined, those from investment and household consumption remain high, whereas government consumption grows more slowly due to tighter fiscal control.

2. From 2007 to 2020, $Fy$ was the main driver of CO₂ emissions in Vietnam. Emission intensity (⁠$FB$⁠) lost effectiveness over time and turned positive in 2016–2020 due to the persistence of carbon-intensive industries. Economic structure (⁠$FMa$⁠) had mixed effects on CO₂ emissions. It reduced emissions in 2007–2012 and 2016–2020 thanks to efficiency gains in sectors like C22, C23, C14 and C28, with the latter period also reflecting temporary reductions from COVID-19 disruptions. In contrast, $FMa$ increased emissions in 2012–2016 due to infrastructure-driven growth and reliance on traditional technologies in C14 and C28.

3. Household consumption drives emissions in C15 and C23, while government consumption impacts C14 and C15. There are clear disparities in household emissions: high-income households emit more due to higher consumption; low-income households emit less, focusing on essentials. Notably, rural households emit more than urban ones due to larger household sizes and higher aggregate demand. Investment triggers high emissions in C4, C10, C15, C23, and C25; inventory increases emissions in C22, C26, and C27; and exports cause the largest emission rises in C15, C23, C27, and C29.

4. The COVID-19 pandemic had both immediate and potential long-term impacts on Vietnam’s final consumption and CO₂ emissions. In the short term, disruptions to production and supply chains slowed the growth of final demand, and emissions from inventory declined significantly as firms turned to stockpiles. However, residential electricity use and related emissions increased during lockdowns. Over the long term, the pandemic may accelerate structural shifts toward digitalization, energy efficiency, and more sustainable consumption behaviours. Nevertheless, without well-designed green recovery policies, post-COVID investments in carbon-intensive sectors could reinforce existing high-emission trajectories.

To align its emission trajectory with the Net-Zero 2050 target, Vietnam needs a comprehensive package of solutions:

First, Vietnam should boost green education, responsible consumption, and low-carbon planning; scale renewables with storage and smart grids; create green finance for solar and batteries; and decarbonise industry via renewables, cogeneration, and waste-heat recovery. Strengthening domestic demand, high-value services, and local supply chains, particularly by SME support, will build resilience and reduce export dependence.

Second, for household consumption, the government should introduce progressive electricity taxes on excess use and provide 0%-interest loans for energy-efficient appliances; subsidise 30% of rural rooftop solar (C23); mandate carbon labels and rebates for recycled steel and green cement in home renovations (C15); extend microcredit and clean-energy grants to low-income families; and bolster domestic green-steel and cement value chains with deep-processing incentives, export limits, risk insurance, and SME credit.

Third, in public spending, especially for C14 and C15, the state must require green procurement of low-carbon materials and technologies, prioritise sustainable transport, green logistics, and digital systems, and embed carbon-intensity checks into all project approvals.

Fourth, investment rules for C4, C10, C15, C23, and C25 should redirect capital toward recycled materials, renewables, and green buildings; mandate emission assessments for major BOT/PPP projects; and back them with tax breaks and green-credit schemes.

Fifth, to curb inventory-related emissions in C22, C26, and C27, policies should promote lean production, digital logistics, and routine carbon audits to eliminate overproduction and storage waste.

Sixth, for exports, Vietnam should set carbon-performance standards for C15 and C23; secure at least 30% renewables via power-purchase agreements; deploy electric trucks (C27) and 5% sustainable aviation fuel (C29); require sustainability reports with green-supply-chain labels; enable green export bonds; and invest in smart grids, energy storage, and green-skills training.

Seventh, to manage episodic emission spikes from final consumption, deploy real-time monitoring with adaptive carbon pricing, automatic “trigger” thresholds, and a low-carbon toolkit (circular-economy measures, flexible production) underpinned by data-driven policymaking. Moreover, COVID-19’s demand shocks highlight the need for green recovery measures, as post-pandemic investments may otherwise lock in high-emission pathways.

Future research should employ causal-inference techniques (e.g. difference-in-differences, instrumental variables) to quantify the contributions of household consumption, investment, and exports to sectoral CO₂ emissions; rigorously evaluate carbon taxes, emissions-trading schemes, and green procurement in major emitters; and assess how trade agreements (CPTPP, EVFTA, RCEP) reshape embodied carbon in global value chains. Building on our mismatch results, studies must also conduct targeted sectoral case analyses, explore nonlinear interpolation for outlier years, and develop dynamic, time-series SDA frameworks to enhance both generalizability and methodological resilience.

This paper forms part of a special section “Sustainability: A Journey for Better Future in Developing Countries”, guest edited by Prof Louis T.W. Cheng.

The supplementary material for this article can be found online.