Drivers of household carbon footprints across EU regions, from 2010 to 2015

Urban regions are responsible for a significant proportion of carbon emissions. The carbon footprint (CF) is a practical measure to identify the responsibility of individuals, cities, or nations in climate change. Numerous CF studies have focused on national accounts, and a few combined consumer consumption and global supply chains to estimate additionally detailed spatial CF. However, the drivers of temporal change in detailed spatial CF are largely unknown, along with regional, spatial, and socioeconomic disparities. Here, we uncovered the drivers of changes in household CFs in EU regions, at the finest scale currently available, between 2010 and 2015. This study mapped the household CFs of 83 macro-regions across 27 EU nations and identified the driving factors underlying their temporal change. We combined multi-regional input-output tables and micro-consumption data from 275 247 and 272 045 households in 2010 and 2015, respectively. We decomposed EU regional CF, employing structural decomposition analysis, into five driving factors: emission intensity, supply chain structure, population, per capita consumption, and final demand share. For a deeper assessment of changes in the contribution of consumption patterns, we further categorized the regional CF into 15 factors, including 11 per capita consumption categories. We found that household CF drivers vary depending on region, population density, income, and consumption patterns. Our results can help policymakers adopt climate policies at the regional level by reflecting on the residents’ socioeconomic, spatial, and consumption conditions, for further ambitious climate actions.


Introduction
The transition to the net-zero carbon emissions is one of the most important challenges for nations, cities, companies, and households. From Kyoto, Paris, to Glasgow, countries have become signatories to climate change agreements, to mitigate global greenhouse gases (GHGs) emissions. Nonetheless, GHG reduction has made limited progress, and national climate pledges are insufficient to meet the Paris agreement GHG reduction targets (Gore 2021, Lamb et al 2021). To achieve net-zero carbon emissions, action from not only the national government but also various stakeholders, such as cities and individuals, is vital to reduce emissions. Carbon footprint (CF) has been used to identify the responsibility of nations, cities, or households for global emissions by linking individual consumption to the production of goods and services (i.e. commodities), through global supply chains (Lenzen and Peters 2009, Kennedy et al 2014, Chen et al 2020, Wu et al 2020. CF assessment can reveal the consumption activities that have contributed to global emissions along with their sources, which is necessary for building climate policies on the demand side (Creutzig et al 2018).
Household consumption was identified as the most dominant contributor to the CF, among the final demands for most developed nations. In particular, household consumption in cities is responsible for 70%-80% of the global CF ( UN Habitat 2011), with 18% contribution from 100 major cities (Moran et al 2018). Therefore, household CF has been the main concern in consumption-based carbon reduction. Recent studies underline the inequalities in household CF, with greater intensity of disparities between specific groups than that between countries Middlemiss 2021, Wagstyl et al 2021). Notable early studies have revealed potential inequalities in the carbon consumption profile among individuals, groups, and regions (Lenzen and Peters 2009, Larsen and Hertwich 2010, Jones and Kammen 2011. Further studies have clarified these disparities based on affluence (Girod and Haan 2010, Golley and Meng 2012, Kennedy et al 2014, Ottelin et al 2018a, Ivanova and Wood 2020, Clement et al 2021, Gore 2021, CF levels (Ottelin et al 2018a, Ivanova andWood 2020), environmental consciousness (Wu et al 2016, Sun et al 2018, religion (Lee et al 2021), or degrees of urbanization (Jones and Kammen 2014a, Ottelin et al 2019b, Hachaichi and Baouni 2021, Ivanova and Middlemiss 2021, Sun et al 2021, Cheng et al 2022, Connolly et al 2022. Empirical studies have indicated that these factors significantly influence household CF (Minx et al 2013, Long et al 2017, Song et al 2019, Shigetomi et al 2021. In addition, because these factors may have regional variance, a finer measurement of spatial CF (i.e. the city's CF) is required to discuss climate policies with greater specificity. While most studies on household CF have focused on the national level, several studies have narrowed it to city-level CF (Jones and Kammen 2014a, Ivanova et al 2017, Moran et al 2018, Kanemoto et al 2020. However, these studies illustrated a snapshot of the CF and did not clarify the effect of these factors on the temporal changes in CF. Structural decomposition analysis (SDA) is a methodology used to assess the underlying drivers of temporal changes in carbon emissions  based on the contribution of physical and economic determinants, such as emission intensity and population . Combined with input-output models (Miller and Blair 2009), SDA provides practical modes for examining sectoral details (Yamakawa and Peters 2011), such as that achieved in Australia (Wood 2009), China (Meng et al 2022), the EU (Duarte et al 2021), Japan , Korea (Lim et al 2009), and the US (Feng et al 2015). However, SDA-related studies have implemented national boundary investigations and not further than the regional level in the EU. Furthermore, clarity is lacking on the exact consumption activities that have contributed to the CF changes in the EU over time, although several studies (see Shigetomi et al 2019, Song et al 2022 revealed that the change in consumption patterns can be a major contributor to emission changes. Considering the municipalities' target-specific climate policies (Ottelin et al 2018b), identifying detailed regional contributors, including consumption categories, is increasingly required.
Research gaps exist between regional-level CF studies and CF decomposition analyses, as their combinations have not been studied. Numerous studies have attempted to measure and mitigate the carbon emissions from cities owing to their significant contribution to this problem (C40 2018, CDP 2020, Cities 2020, ICLEI 2020, UN 2020). Urban consumers have largely contributed to global emissions due to their affluent consumption (Minx et al 2013, Kennedy et al 2014, Wiedmann et al 2016, Creutzig et al 2021, Wiedmann and Allen 2021, although the high population densities and efficient facilities in the cities are advantageous for carbon mitigation (Heinonen et al 2013). Furthermore, several cities have experienced rapid demographic shifts (UN 2018(UN , 2019 and spatial expansion (Seto et al 2012), resulting in rapid lifestyle and population changes that affect urban emissions. Consequently, cities have been actively pursuing policies for net-zero carbon emissions. However, most regional CF studies have not accounted for these social changes, and SDA studies lack in such details within national boundaries. Almost all previous findings on the regional CFs have limited our understanding to specific years, without examining how varied circumstances have contributed to the CF change. Nation-level SDA studies have not revealed the disparities among different spatial or socioeconomic groups. Recent studies have indicated a regionally specific relationship between population density and CF (Ottelin et al 2019a. Household CF estimation can reveal inequalities within the CFs of different socioeconomic and demographic groups (Howell 2013, Büchs 2014, Shigetomi et al 2014, Kanemoto et al 2019. Decomposing regional or group-specific CFs can disclose the underlying driver contributions to the temporal CF change in various individuals, enabling target-specific policies, to meet our ambitious climate action goals. This study had three aims: first, to estimate the temporal changes in regional CFs across the EU in as much detail as currently possible; second, to identify the underlying drivers of CF changes over time through the SDA; and third, to uncover the regional, spatial, and socio-economic disparities. We estimated the CF of 275 247 and 272 045 households in 2010 and 2015, respectively, across 83 NUTS1 (major socioeconomic regions, and subdivisions of the countries) regions in 27 EU nations 5 and calculate regional CF accounts through household CFs, whereas previous studies allocated national-level CF to gridded data using population and income data without regionscale datasets (see Moran et al 2018). We present the first regional and group-specific drivers of temporal CF changes by NUTS: densely and sparsely populated areas, wealthy and deficiency, and high-and low-CF consumers. Furthermore, we indicate the contributions of shifting consumption patterns to regional CF changes over time. These novel outcomes reveal uneven responsibilities within regional, spatial, economic, and social disparities, contributing to greater exhaustive coverage and precise climate policy targets.
We addressed the following research questions, encompassing EU household CF: 1. How has household CF changed between 2010 and 2015? 2. What are the main drivers contributing to CF changes over time, and which consumption activities are responsible? 3. How do underlying drivers differ among other regional, spatial, and socioeconomic groups?
Section 2 describes the methodology and data used in this analysis. Section 3 presents the results of the study. Section 4 presents the limitations of our analysis. Finally, section 5 conveys avenues for future empirical and methodological work.

Methods
Regional or group-specific CF assessments have grown in response to the demand for more suitable target-specific climate governance. These assessments distinguish between the carbon emissions associated with different regions, individuals, or social groups. Integrating micro-consumption data into global supply chains has provided meaningful information on groups of districts (Lee et al 2021), incomes (Ottelin et al 2018a), and health conditions (Ivanova and Middlemiss 2021), which are not available in countrylevel CF estimations. Regardless of the growing policy demand for detailed accounts, developing groupspecific CF accounts remains challenging. Within this study, we constructed CF accounts by NUTS1 region, 5 Eurostat provides microdata on household expenditure on goods and services for the reference years 2010 and 2015, harmonizing survey data from 27 EU nations. The data convey information on household consumption at the commodity level for each 196 (in 2010) and 299 (in 2015) items in a consistent framework across EU countries. Therefore, this study tracked temporal changes over a five-year period. The 27 countries include Belgium, Bulgaria, Croatia, Cyprus, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, and the United Kingdom. population density, income levels, and emission levels across EU nations, analyzing micro-consumption data from 275 247 and 272 045 households in 2010 and in 2015, respectively.
In this study, we employed a personal CF assessment approach identified by Heinonen et al (2020Heinonen et al ( , 2022, which allows a thorough estimation of the individual consumption patterns across detailed expenditure categories by aggregating personal CF into group accounts. In contrast, the areal CF approach considers the total expenditure within a sample group. The CF assessment utilizes monetary multi-regional input-output table (MRIO) databases that integrate economic and emission flows across global production and consumption systems. We integrated household consumption data into the Eora (2022) MRIO database by constructing concordance tables in this study, as described in detail in section S6. The integration involved the linking 196-299 consumption categories in household budget survey (HBS) 6 to 26-511 commodity and service sectors in the Eora (2022) database developed by Lenzen et al (2012Lenzen et al ( , 2013. We estimated the CF of each group within the EU households by integrating and harmonizing the HBS data with a global supply chain database, as described in section S13. To unveil the drivers of CF changes from 2010 to 2015, this study utilized SDA. SDA enumerates the contribution of the emission-related factors toward CF change over time, linking altered situations and consumption shifts, using MRIO tables. This approach incorporates the advantageous features of the MRIO database with detailed sectoral resolution . The Eora (2022) database offers benefits for the application of SDA with its high resolution . Although there is no unified methodology to conduct decomposition, and develop emission indicators over time, we implemented a decomposition method that averages all firstorder decompositions to avoid biases depending on the decomposition modes (see Dietzenbacher and Los 1998, Hoekstra and van der Bergh 2002, Rørmose and Olsen 2005, Yamakawa and Peters 2011, while it constrains the number of factors owing to its intensive calculation (Su and Ang 2012), as detailed in section S14. 6 Eurostat's HBS provides detailed data on household consumption of food, electricity, gas, consumable or durable goods, transportation, education, medical care, sanitation, entertainment, etc in 196 (in 2010) or 299 (in 2015) commodity and service sectors, such as bread, milk, natural gas, heat energy, non-durable articles, furniture, diesel, primary education, hospital services, water supply, sewage collections, television subscription, wireless telephone service, or accommodation services. The appended file 'Sup-plementary_data.xlsx' (tab 'SI_EU_HBS_commodities') conveys lists of commodities and services for 2010 and 2015. Using the microdata, we estimated the household CF that reflects domestic consumption.
The data sources utilized in our analysis, the HBS microdata (Eurostat 2016, 2021), Eora MRIO database (Eora 2022), and Gridded Population of the World database (CIESIN 2018), are described in sections S2, S3, and S4. We bracketed the HBS samples for each country by NUTS1, population density, household net income, and household CF thresholds, to indicate the drivers within the different spatial and socioeconomic groups. Section S5 describes this grouping in greater details.

EU household CF profiles
This section illustrates the observations across 199 consumption sectors of 275 247 households in 2010, and 299 of 272 045 in 2015 across 27 countries. Within this section, the CF is expressed in tons of CO 2 /capita/year (hereafter, 'ton CO 2 /capita').

Spatial and temporal differences
The NUTS1-level CF maps for 2010 and 2015 are illustrated in figures 1(a) and (b), which display the latest spatially explicit CF for estimating direct and embedded carbon emissions associated with the EU household consumption. The map in figure 2 illustrates the gaps from 2010 to 2015. The most prominent of these are the growing disparities within the regional carbon consumptions when the CF distributions in 2010 and 2015 are compared. Meanwhile, CF changes from 2010 to 2015 present a promising consequence for overall CF reduction. Our analysis estimated the mean CF of individuals in 26 EU countries as 8.98 ton CO 2 /capita for 2015, which is significantly lower than the 11.29 ton CO 2 /capita in 2010. Denmark (5.80 ton CO 2 /capita decrease), Cyprus (5.13 ton CO 2 /capita decrease), and Makroregion południowo-zachodni NUTS1 in Poland (4.8 ton CO 2 /capita decrease) have achieved remarkable individual carbon reductions, while Isole NUTS1 in Italy indicated an increase (1.53 ton CO 2 /capita risen). Additional maps and the database pertaining to this study can be found in section S1 and are available online at https:// city.spatialfootprint.com/#eu for research and policy applications (Kanemoto et al 2023).

Decomposing household CF by region and group
This section summarizes the quantified contributions of physical and economic determinants to the CF changes over time, within regional, spatial, and socioeconomic groups across 25 EU nations 7 .

Key drivers of regional carbon mitigation
We quantified the contributions of driving factors to the CF changes over time in each NUTS1 region by decomposing the regional CF into five factors: emission intensity, supply chain structure, population, per capita consumption, and final demand share. For an additionally specific assessment, our estimation included the contribution of each categorized consumption change by decomposing the CF into 15 factors 8 .
Across the 64 NUTS1 regions, reduced per capita consumption primarily contributed to carbon mitigation, followed by supply chain structure and emission intensity, while changes in population and final demand share had relatively minor contributions. This indicates that moderate consumption primarily drove carbon mitigation across the NUTS1 regions in the EU, especially in Greece (contributing to a 34.3% decrease in 2010 overall CF, hereafter −34.3%), Sur NUTS1 in Spain (−26.9%), Nord-Ovest NUTS1 in Italy (−25.9%), and Sud-Ouest NUTS1 in France (−24.6%). Meanwhile, consumption growth in Latvia (contributing to a 7.1% increase in 2010 overall CF, hereafter +7.1%) and Lithuania (+5.8%) contributed to the CF increase. The contribution of change in emission intensity varied greatly among EU nations. Intensity changes significantly contributed to the CF decrease in Hungary regions (−15.9 to −17.0%), Malta (−14.2%), and Lithuania (−13.3%), contrary to the contributions to a CF increase in Greece (+19.8%), Italy (+6.3 to +7.4%), and Spain (+5.0 to +5.5%). Changes in supply chain structures (i.e. product recipe change) contributed significantly to the overall carbon reduction across 25 nations, while population change (reflected partially by demographic change) and final demand share (affected by the importing profile) moderately drove the CF changes between 2010 and 2015. Contributors to the temporal CF change differed significantly depending on the NUTS1 region, although we observed overall trends in the contribution of physical or economic factors. For the emission intensity and supply chain structure factors, we applied national accounts, which differ from NUTS1 level accounts, for other factors that probably cause minor differences between regions within the same country.
Among the per capita consumption categories, the change in private transport consumption exhibited the largest contribution to CF change across NUTS1 regions. The consumption of food, electricity, and other services also significantly drove the CF mitigation in the NUTS1 regions, followed by other energy, gas, durable goods, consumable 8 The fifteen factors are emission intensity, supply chain structure, population, final demand share, and eleven categorized per capita consumptions of food, energy, gas, other energy, public transport, private transport, medical care, education, consumable goods, durable goods, and other services. goods, and public transport consumption. Medical care and education consumption contributed the least to CF change. Even in the Germany and Italy NUTS1 regions, which indicated overall growth in private transport CF, the lowered consumption in that category contributed to CF reduction. Across all the 64 NUTS1 regions, shifting food consumption contributed to CF mitigation, especially in Luxembourg (−6.0%), Região Autónoma dos Açores in Portugal (−5.8%), and Croatia (−4.2%). Regional differences were also observed. In Romania, changes in the per capita private transport consumption in Macroregiunea Patru NUTS1 (−14.4%) significantly contributed to the regional CF reduction, in contrast to Macroregiunea Trei NUTS1 (+3.8%). Despite the overall low contribution of medical care consumption, it raised the regional CF of Macroregiunea Patru NUTS1 in Romania (+2.5%), while reducing that of Åland NUTS1 in Finland (−1.9%).
In summary, we observed that reductions in private transport and food consumption primarily drove carbon mitigation across the EU NUTS1 regions. These could be in response to severe climate policies targeting the transportation modes and dietary habits, such as Sweden's carbon tax on transport fuels (Andersson 2019), the Netherlands' CO 2 -differentiated vehicle tax (Dineen et al 2018), and Spain's carbon tax on animal-sourced foods (Forero-Cantor et al 2020). The considerably different contributors across NUTS1 regions indicate the need for regional and target-specific climate policies. For example, Sweden's relatively low per capita energy-related CF (electricity, gas, and other energy) compared with that of its neighbors can be explained by policies and practices, such as promoting the installation of enhanced insulation or heating systems in new buildings, as highlighted by Ottelin et al (2018b). Detailed results on the contributors in the 64 NUTS1 can be found in sections S1 and S19.

Population density differences
The outcomes of our assessment indicate that the contribution of individual consumption to carbon emissions varied considerably with population density, as illustrated in figure 3. These results have important implications pertaining to the population density groups, as several arguments have been forwarded on the influence of population density on carbon emissions, without reaching a compromise (see Shammin et al 2010, Jones and Kammen 2014b, Moran et al 2018, Ottelin et al 2019a. Within this study, we quantified the contributions of factors in population groups dwelling in sparsely, intermediate, and densely populated areas 9 . Detailed results are presented in sections S1 and S19. Spatial differences are a significant factor in CF in several countries, including Belgium, Bulgaria, Cyprus, Estonia, Finland, Hungary, Ireland, Latvia, Lithuania, Luxembourg, Spain, and Sweden, however insubstantial in the Czech Republic, Greece, and Italy. Within the EU boundary, the contribution of individual consumption volume to emission change is comparable among densely (−16.9%), intermediate 9 Following Eurostat (2021)'s definition, we define densely, intermediate, and sparsely populated areas as areas of 'at least 500 inhabitants/km 2 ,' 'between 100 and 499 inhabitants/km 2 ' , and 'less than 100 inhabitants/km 2 ,' respectively. The contribution of population changes in densely, intermediate, and sparsely populated areas varied by region in the EU. Population growth in densely populated areas contributed to increased carbon emissions, as in Luxembourg (+16.8%), Sweden (+8.3%), Hungary (+5.5%), Belgium (+5.5%), Finland (+5.1%), and Denmark (+3.7%). While, population changes in sparsely populated areas contributed to carbon reduction in Finland (−6.1%), Denmark (−1.8%), Luxembourg (−1.0%), and Hungary (−0.4%), but not for Belgium (+1.5%) and Sweden (+0.1%). In several intermediate populated areas, it contributed significantly to a decreased emission in Latvia (−33.3%), Bulgaria (−13.7%), and Hungary (−8.8%), in contrast to the increased emission in Finland (+24.8%), Cyprus (+9.7%), and Luxembourg (+6.6%).
Our results revealed that the driving factors for emissions change differ between rural (sparsely populated), and urban (densely populated) areas, although less than those between income groups. Population shifts between rural and urban areas can affect the regional CF. This outcome highlights the necessity of rural-or urban-specific climate policies in the context of demographic movements, which national accounts have not captured.
In high-income groups, a decrease in the wealthy population contributed to an overall decline in CF in the EU (−11.4%, 51 megatons CO 2 decrease). In contrast, a growth in wealthier populations contributed significantly to an increased CF, especially in the high-income groups of Estonia (+986.0%), Malta (+256.9%), Hungary (+118.0%), Slovenia (+80.3%), and Romania (+53.7%). Overall, the disparities between income level groups exhibited greater extremes than those between NUTS1 regions or spatial groups by population density. The contribution of each factor in different nations is described in sections S1 and S19.

CF differences
According to our results, several households were incorporated into the lower-emission groups, with considerable shifts from middle-and high-emission households, while Bulgaria displayed fewer low-CF households. Across 25 nations, the contribution of each of the five factors to the change in CF over time varied enormously by CF emission level and country. We observed that the affluent consumption of high-CF groups in certain countries had diminished the carbon mitigation, such as in Latvia (+41.9%), Estonia (+27.1%), Bulgaria (+25.8%), Portugal (+21.9%), and Denmark (+17.8%). Meanwhile, the lowered per capita consumptions of high CF groups in Slovakia (−22.5%), Italy (−15.2%), Slovenia (−15.1%), Luxembourg (−13.1%), and Germany (−9.3%) contributed to CF reductions. This study set the CF brackets of 8.0 ton CO 2 /household and 50.0 ton CO 2 /household, to divide households roughly into low-(bottom 10%), middle-(middle 80%), and high-CF (top 10%) emitting groups, as detailed in section S5.

Discussion
We estimated the household CF for 83 NUTS1 regions across 27 EU nations on an annual basis for 2010 and 2015, and conducted the first regional-scale SDA for 64 NUTS1 regions in 25 countries. This study confirmed that 63 NUTS1 regions, except one, accomplished reduced individual CFs. Our assessment revealed that individuals in the EU reduced their average personal CF by 20.5%, from 11.29 ton CO 2 /capita in 2010 to 8.98 ton CO 2 /capita in 2015. The SDA outcomes indicate that reduced per capita consumption is the primary contributor to regional carbon mitigation, particularly in the private transport, food, and energy sectors, followed by decreased carbon intensity. This suggests climate policies that reconcile consumption reduction with carbon intensity declines rather than relying solely on the latter, such as carbon-reducing technologies or renewable grids.
We observed that the drivers of CF changes over time depend significantly on regional, spatial, and socioeconomic contexts, suggesting the need for regional or group-specific climate policies. In Finland, changes in the per capita consumption of the Manner-Suomi NUTS1 and Åland NUTS1 had opposing contributions, indicating the necessity for a more intensive review of consumption in the Åland region. In Romania, changes in private transportation consumption lowered the regional CF of Macroregiunea Patru while elevating it in the neighboring Macroregiunea Trei. In Luxembourg, each population change in densely and sparsely populated areas contributed contrastingly to CF changes. This suggests linking demographic shifts to climate policies for additionally comprehensive climate actions. The emission intensity factor significantly drove the CF change in regions of Hungary, while that in Portuguese regions was primarily affected by the supply chain structure. Each indicates the need to review the on-site production technologies or manufacturing recipes. Per capita consumption changes contributed greatly to the CF reduction in Greece and Sur NUTS1 in Spain while increasing the CF in the Latvian and Lithuanian regions. This implies that consumption changes do not evenly contributed to carbon reduction across EU regions, suggesting that additionally intensive alterations in consumption have to be implemented in these regions.
Meanwhile, the SDA results revealed there are greater extensive disparities among socio-economic groups, wealthy and deficient, or high-and lowemission households, than among the regional or spatial differences. An increased consumption from affluent groups in several nations, such as Cyprus and Greece, contributed to CF growth, while consumption shifts in middle-and low-income households in the same nations contributed to a decrease in CF. Furthermore, notable gaps were observed in specific consumption categories, such as the elevated medical carbon consumption by high-income households. Major climate policies have concentrated on urban habitats, mainly on food, electricity, and transportation consumption. However, we suggest regional climate policies encompassing rural households and other consumption categories, such as medical care or durable commodities, which greatly contribute to CF changes in specific regions or groups.
The prominent responsibilities in the service sector of high-income households in several nations imply that service consumption has great potential to abate household emissions. Alongside restaurant and cafe consumption in the service sector, the expansion of internet-driven media consumption has raised the regional CF significantly, in several nations such as Bulgaria, Romania, and Lithuania. Notably, national and municipal climate policies need to reflect the CF arising from carbon-consuming activities to fulfill the national pledges for the Paris Agreement.
The results of our assessment should be understood within the limitations concerning data availability and quality, methodological assumptions, and the research scope. The HBS microdata underpinning our analysis screened the consumption data of several outermost wealthy households, following Eurostat's anonymization policy, which potentially lowers the average individual consumption. We scaled the household consumption in HBS to satisfy the national consumption accounts of Eurostat. The resolution of the Eora MRIO database varies between 26 and 511, and several developing countries have 26 commodity and service sectors. We linked the HBS consumption sectors to the Eora sectors, in concordance with sectoral relevance. In addition, Eora does not supply subnational MRIO accounts, such as those provided by Lenzen et al (2017), Wakiyama et al (2020), Zheng et al (2021). This limited our SDA outcomes, which integrated the national-level MRIO tables with regional-level consumption microdata. Furthermore, it applies the national accounts for the emissions intensity and supply chain structure factors, which differ from the NUTS1 level for the others. Nevertheless, the Eora database provides the most detailed resolution of global supply chains worldwide. In converting MRIO accounts into constant prices, there were mismatched sectoral details between the Eora database and HICP (employed for the conversion), while both apply the classification of individual consumption by purpose. This study focused on household consumption and excluded the carbon emissions related to government and capital consumption. Most importantly, the explanatory power of our estimation was determined by the scope of HBS. The HBS microdata are available only for the reference years 2010 and 2015, which limits the temporal scope of our estimations to five years. Eurostat has harmonized household surveys in various countries with significant divergence. Although this variation can affect the consistency of our analysis, Eurostat's harmonized survey database delivers a valuable and diverse cross-cutting of household livelihoods across the EU.

Conclusion
This study aimed to identify the underlying drivers of CF changes over time in as much detail as possible, revealing regional, spatial, and socioeconomic disparities. Notwithstanding these limitations, this study provides the latest regional accounts of the household CFs of 27 EU nations from 2010 to 2015, drivers of CF changes over time, and the disparities within regional, spatial, and socioeconomic groups. Our estimation suggests the need for regional and group-specific climate policies concerning new demands. It provides a promising signal for attaining regional carbon neutrality that includes all societies and lifestyles. Based on our findings, a sustainable future requires decisive target-specific climate actions in the EU and across the world.
Future work should be additionally extensive, covering a greater extended period and scope of surveys to identify trends and disparities in CF changes. Understanding what leads to changes in CF remains a future challenge. Discovering which policies and practices, such as promoting public transport or energy-saving buildings, contribute to carbon mitigation can provide practical and effective guidance to local policymakers. It is necessary to assess the contribution of changed fuel sourcing by examining the consumption data in the physical unit, which is impossible with the monetary unit-based microdata (see Hoekstra et al 2016).

Data availability statement
The data that support the findings of this study are openly available at the following URL/DOI: https:// city.spatialfootprint.com/#eu.