Examining the reasons for changes in buildings’ energy consumption in the United States, China and the European Union

Energy & Buildings 332 (2025) 115461 Available online 11 February 2025 0378-7788/© 2025 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/). Examining the reasons for changes in buildings’ energy consumption in the United States, China and the European Union M. Gonz´ alez-Torres a,b,* , L. P´ erez-Lombard b , E.L. Clementi c , J.F. Coronel b a European Commission Joint Research Centre Directorate B – Fair & Sustainable Economy Seville Spain b Grupo de Termotecnia Escuela Superior de Ingenieros University of Seville Spain c European Commission Joint Research Centre Directorate C - Energy, Transport and Climate Ispra Italy ARTICLE INFO Keywords: Buildings’ energy use Building sector Activity drivers Energy-use intensity Decomposition analysis Logarithmic Mean Divisia Index (LMDI) ABSTRACT Buildings are responsible for one third of global operational energy consumption and greenhouse gas (GHG) emissions. Addressing their impact requires the development and monitoring of effective policies, supported by detailed and costly data on building stock and energy use as well as their corresponding analysis. The paper proposes a pyramidal approach to decompose buildings’ energy use into drivers —activity, structure, and efficiency— considering factors like population, floor area, urbanisation, building size, occupancy and climate. Energy-use intensity measures efficiency, while shifts among the residential and tertiary subsectors are captured as structural impacts. The relevance of the methodology is underscored by its potential to assess and quantify the causes of energy consumption changes, guiding policy-making. Applying this approach to China, the United States (US) and the European Union (EU), the paper criticises the lack of data, disaggregates energy consumption changes, outlines policy implications and validates the methodology’s added value. The analysis reveals the increased floor area as the primary driver of rising energy consumption over the past two decades (contributing to changes by 9% in the US, 24% in the EU and 97% in China). This may be reduced by managing urbanisation rates and compensated by an improvement in efficiency. While this has been sufficient to stabilise consumption in the EU, a slight rebound is observed in the US due to the increase in population and in the demand for buildings per capita. In China, the urbanisation trend behind the rise in energy demand is approaching EU levels, highlighting the importance of mindful actions to ensure the sustainability of future expansion. Despite the limited geographical coverage, this study provides a pertinent analysis of almost half of the building energy consumption in the world (China, the US and the EU), offering insights into the sector’s current state and directions for future policy development. 1. Introduction The impact of the building sector continues to be a matter of concern. On the one hand, operational GHG emissions in buildings accounted for 30 % of the global consumption figures [1]. Both new constructions (that should be converted to fossil-fuel-free and zero-emission buildings 1 ) and the existing stock (that should be energy-retrofitted at a faster pace) need to be addressed to mitigate the issue and halt the increasing impact of the sector [2]. On the other hand, embodied emissions, which include those for the buildings’ construction, have risen to 10 % worldwide, mainly due to the growing demand for materials [1]. Together making up 40 % of global emissions, the reduction of buildingrelated emissions is therefore not only necessary but urgent. As the main driver of global emissions, energy consumption should be carefully considered as a means of reducing impacts [3,4]. The growing energy demand for the operation of buildings has almost equalled that of the industrial sector, representing 28 % of the energy consumed globally in 2022 [5]. This rises to 36 % when also considering the energy required for their construction, embodied energy [6], making buildings the highest energy-consuming sector worldwide and * Corresponding author. E-mail address: [email protected] (M. Gonz´ alez-Torres). 1 Note that the term zero-emission building (ZEB) here is used as defined by the energy performance of the Buildings Directive (EPBD) (Directive (EU) 2024/1275): ‘‘zero-emission building’ means a building with a very high energy performance, as determined in accordance with Annex I, requiring zero or a very low amount of energy, producing zero on-site carbon emissions from fossil fuels and producing zero or a very low amount of operational greenhouse gas emissions, in accordance with Article 11 ′ . Contents lists available at ScienceDirect Energy & Buildings journal homepage: www.elsevier.com/locate/enb https://doi.org/10.1016/j.enbuild.2025.115461 Received 13 December 2024; Received in revised form 31 January 2025; Accepted 10 February 2025 Energy & Buildings 332 (2025) 115461 2 evidencing the need for analysis and action. Energy policies in buildings globally reduce the share of operational energy use [7], but there is still much room for improvement. In developing countries, absolute consumption figures more than doubled from 2000 to 2022, increasing at an average annual rate of 1.9 % [5]. The reality is harsh since trends in consumption and emissions reveal that technology and policy are not enough to tackle the problem. A socio-economic view of the problem is necessary. Buildings are constructed to meet the needs of their occupants, so the services demanded depend primarily on human activities. Therefore, the key question could be: why are we increasing buildings’ energy use? The systemic analysis of the reasons why energy is consumed must necessarily consider three aspects: activity, structure and intensity. The first focuses on the phenomena that drive the demand for services, the second explains the impact of different activities or sectors, and the third is used as a measure of efficiency since it links the energy consumed and the service provided [8]. In recent decades, the decomposition into these factors has been widely used to achieve a disaggregated analysis of energy and emissions. The United Nations (UN) and the International Energy Agency (IEA) funded various projects in this field and authors such as L.J. Schipper [9], F. Unander [10] and Y. Kaya [11] were pioneers in explaining the links between energy and economics using decomposition analysis. Over time, decomposition techniques have evolved to become a mature and robust tool. Today, Index Decomposition Analysis (IDA), and in particular the Logarithmic Mean Divisia Index (LMDI) [12], is widely applied for energy analysis, to explain changes and to monitor and develop policies, assisting in decision making. The research and reviews published by B.W. Ang [12–15] have established the terminology and doctrine in this field. When decomposition techniques are applied to the building sector in order to answer the research question above, the literature (whose contributions are thoroughly classified and examined in Section 2) allowed us to identify the following gaps. First, most papers focus on understanding the drivers of individual subsectors (residential, public, commercial buildings), while analyses of the building sector as a whole are scarce. Second, the usual practice in the literature consists of standalone decompositions, while a stepwise approach would allow for a better understanding of the underlying factors driving the consumption growth. Third, the structural effect among residential and tertiary subsectors has never been evaluated. Finally, most publications in the field decompose the energy consumption of a single nation, mainly China, while articles with international comparisons are almost non-existent. This paper aims to fill the gaps above by proposing a pyramidal approach that progressively decomposes the energy consumption of the whole building sector into its main activity, efficiency and structural drivers. Population, floor area, urbanisation, building size and occupancy are used as activity and lifestyle factors, while the energy-use intensity is used to assess the energy efficiency. Additionally, the impacts of climate and structural changes are assessed by applying weather corrections and by distinguishing between the residential and tertiary subsectors. Table 1 defines these indicators, which are further described in Section 3. The proposed approach is applied to relevant nations, providing the cross-country comparison that is missing in the literature. Although the geographical coverage is limited due to data constraints, the study is conducted for China, the United States (US) and the European Union (EU), the highest consumers in the world, meaning an up-to-date analysis of almost half of the global energy consumption in buildings. This allows meaningful conclusions to be drawn: (a) the data availability is critically scrutinised, (b) the reasons for the changes in buildings’ energy use are revealed, (c) the policy implications are outlined and (d) the added value of the methodology is illustrated. Accordingly, the paper is structured as follows. First, the literature decomposing buildings’ energy consumption is reviewed and classified according to the methodological choices made. Then, the pyramidal decomposition method and the data sources for its applicability are described. Next, the drivers of changes in buildings’ energy consumption in China, the US and the EU are assessed and discussed. Finally, the policy implications of the results are outlined and the main conclusions are highlighted. 2. The art of decomposing buildings’ consumption When decomposing buildings’ energy consumption, important methodological choices need to be made in terms of scope (subsectors and types of buildings), activity and efficiency indicators (for the total energy or by end-use), lifestyle aspects (which disaggregate the activity effects), corrections to the efficiency indicator (structure and weather corrections) and data availability. The following subsections describe such methodological choices, allowing us to analyse the contributions in the literature, classify them accordingly (see Supplementary Information, Table SI 1) and identify the gaps previewed in the introduction. 2.1. Building subsectors and building types Buildings are often not recognised as a separate sector. Traditionally, they have been hidden within the large ‘Other’ sector or need to be roughly calculated as the sum of the residential and tertiary sectors (the latter is also referred to as the service sector, commercial and public sector or simply non-residential) [16]. More complex is the subdivision of the building sector in China [17], which combines subsectors (residential, public and commercial) with end-uses (e.g. heating) and locations (rural vs. urban). Buildings’ consumption in Chinese statistics is the sum of Rural Residential (RR), Northern Urban Heating (NUH), Urban Residential (UR) excluding NUH, and Public and Commercial (P&C) excluding NUH. Consequently, the first decision for the decomposition of building consumption is whether the aggregate magnitude is the whole building sector or a part of it, namely a building subsector (residential vs. tertiary) or type (single or multifamily dwellings in residential buildings, hotels, offices, schools, etc. in tertiary buildings). In the literature, this is a main factor for classifying publications (Table SI 1). Most references focus on a single subsector, mainly residential [18–24] but also tertiary [25–28]. Others decompose both, although they do it independently so they do not explain the structural changes between them [29,30]. Some researchers also target one typology of building within one of the subsectors, noting a focus on public Table 1 Description of activity, efficiency and structure indicators for buildings’ energy analysis. Type Indicator Description Unit Activity P Population − A Floor area m 2 B Number of buildings − u Urbanisation, as the floor area per capita m 2 /cap b Buildings per capita, as the inverse of occupancy build/ cap a Building size, as the floor area per building m 2 / build c Climate correction factor, as the ratio of the actual to the weather-corrected energy consumption − Efficiency e B Energy-use intensity of buildings kWh/ m 2 e Bc Energy-use intensity of buildings with the weather correction kWh/ m 2 e Bcs Subsectoral (residential or tertiary) energy-use intensity of buildings with the weather correction kWh/ m 2 Structure s s Structural factor, as the share of the floor area corresponding to the subsector s (residential or tertiary) % M. Gonz´ alez-Torres et al. Energy & Buildings 332 (2025) 115461 3 buildings [31–34]. Research decomposing the building sector as a whole is scarce and only available for China [35–39]. In this paper, a gap in the scope is filled by analysing the whole building sector for the regions with the highest consumption in the world (China, the US and the EU). 2.2. Activity drivers and efficiency indicators The choice of the main activity driver is directly linked to the definition of efficiency, since it is normally measured by the energy intensity (i.e. energy consumed per unit of activity). However, the specificities of the different building types and end-uses make the choice of the most adequate activity and efficiency indicators complex and diverse. For the energy use in residential buildings (E r ), the floor area (A r ) and the number of dwellings (B r ) are the most common activity drivers, and the energy-use intensity (E r /A r ) and the dwelling consumption (E r /B r ) are normally used as efficiency indicators. In the case of energy use in non-residential buildings (E t ), their link to the service sector of the economy suggests that, in addition to the floor area, the economic output of the tertiary sector (G t ) may also be an appropriate measure of the activity, and the energy per economic output (E t /G t ) would then be used as the efficiency indicator. In both examples, the total-energy use is the aggregate magnitude to be decomposed, so it could be referred to as a ‘total-energy approach’ [40]. However, when the energy information is broken down by energy service, an ‘end-use approach’ allows the identification of the most suitable activity driver for each, independently. For instance, it seems clear that the consumption of domestic hot water depends more on the heated volume and the number of occupants than on the floor area of the building, which justifies the use of the energy consumption per litre or per capita as a more adequate efficiency indicator. Similarly, the consumption in appliances depends mainly on the level of equipment (number of devices), suggesting the energy consumption per device as an adequate efficiency measure. Thus, the choice of the activity and the efficiency indicators is interlinked and has implications on how the decomposition analysis is conducted, influencing the additional factors that can be introduced in the equation, as explained in the Section 2.3. In conclusion, the second decision for the decomposition would be the choice of a ‘total-energy’ or an ‘end-use’ approach and its corresponding main activity and efficiency indicators. The literature review (Table SI 1) shows that most of the authors choose a total-energy approach, with the energy-use intensity being the most common efficiency indicator and the floor area the main activity driver [19,20,22,23,26,27,30–32,35,36,38,39]. Moreover, the economic output of the tertiary sector is used in many of the decompositions that specifically target this subsector, then defining the efficiency as the energy use per economic output [25,28,29,33,34,37]. Also, some studies in the residential sector use other indicators such as energy or household expenditures [18,21] to measure the activity, then using the energy consumption per unit of expenditure as the efficiency indicator. Fewer researchers followed an end-use approach [22,24,41–43], allowing the definition of targeted indicators for specific energy services, such as the energy use per capita for hot water or cooking, the energy use per device for equipment or the energy use per conditioned area for space heating. In this paper, a total-energy approach is adopted and the floor area and the energy-use intensity are chosen as the main indicators of activity and efficiency, in line with the methodological choices in previous studies of its kind. 2.3. Lifestyle aspects Lifestyle involves living conditions, behaviour and habits and comprises many underlying effects that cause changes in the activity within the building sector. For instance, the growth of the residential floor area (A r ) may be driven by an increasing population (P), larger housing sizes, defined as the area per dwelling (A r /B r ), and/or a reduced household size, defined as the people per dwelling (P/B r ). Then, despite the floor area being chosen as the main activity indicator, additional aspects could be analysed by further decomposing it, for instance as: Ar=P⋅Br P⋅Ar Br (1) In the equation above, in line with other reference sources such as the IEA [43], the floor area is the main activity measure that builds the efficiency indicator (energy-use intensity). The population and the housing and household sizes are used to assess socio-economic effects. They can be referred to as lifestyle factors [40] and appear when an activity indicator is decomposed to explain the reasons for its changes. Analogously, the tertiary floor area (A t ) may be decomposed to reveal other effects highly related to the lifestyle of the society, such as the population, the economic output of the tertiary sector per capita (G t / P), and the area productivity, defined as the area per tertiary economic output (A t /G t ): At=P⋅Gt P⋅At Gt (2) When a total-energy approach is adopted to decompose the whole building sector, some authors such as Weiguang et al. [35] and Chen et al. [38] choose a residential perspective (decomposing the area into the population and the area per capita), while others such as Cui and Xia [39] choose a tertiary perspective, introducing the Gross Domestic Product (GDP) per capita and the area productivity as lifestyle factors. Consequently, the third decomposition consists of choosing the lifestyle aspects to be disentangled, i.e. the underlying effects within the main activity indicator. This choice makes a difference to the number of factors analysed and their definitions. In the literature, most studies investigate between two and three lifestyle factors, with a maximum of five being identified in Wang et al. (2022) [33]. In addition to the housing and household sizes, other socioeconomic effects studied in the residential sector are the population, the number of buildings, the urbanisation (calculated as the area per capita), the income or living expenditure per capita or the share of energy expenditure to total expenditure. In the case of studies focused on the tertiary sector, other factors are the number of employees, the hours worked or the labour productivity (volume of economic output produced per unit of labour). A detailed and more complete classification of the indicators assessed by the publications reviewed is described in Table SI 1. This paper differs from the existing ones in the fact that the hierarchical decomposition developed allows for the assessment of up to five lifestyle indicators in a sequential way, providing then the most stepwise analysis and facilitating the drawing of conclusions. These are the floor area, the population, the urbanisation, the buildings per capita (as the inverse of occupancy) and the building size. 2.4. Corrections to the efficiency indicator Like the decomposition of the main activity driver into lifestyle factors, the efficiency indicator can be disaggregated into underlying factors. The energy intensity in buildings is influenced by how the sector is structured. For instance, assuming that hospitals consume more energy than shops (so being more energy-intensive), a higher number of hospitals in the total stock would raise the total energy intensity of the sector. Since this should not be interpreted as a worse efficiency, the impact of structural changes should be isolated. The literature review (Table SI 1) shows that not all authors include structural factors in their decompositions, as it complicates the calculations and the application of the LMDI. The authors have performed structural analyses by end-use [18], by fuel [22,25,28,29,38], by M. Gonz´ alez-Torres et al. Energy & Buildings 332 (2025) 115461 4 building type (dwelling type [19,22] or tertiary typology [28]), or by location (rural vs. urban [30,35,36] or regional analysis [21,22,29]). Nevertheless, no study focusing on the structural shifts among building subsectors has been found. This paper aims to also fill this gap by examining the impact of changes in the contributions of residential and tertiary buildings to the overall floor area in the major economies within the scope. Similarly, the energy intensity is affected by the weather. Low temperatures can increase the use of space and water heating in a building, which does not mean a decrease in efficiency but an increase in demand. Moreover, other weather-dependent conditions, such as daylight, temperature and humidity, can have an impact on the use of certain equipment (lamps, refrigerators, dryers, etc.) and on the number of hours indoors. Thus, the response to weather changes is also sometimes analysed in the literature [19,22,27] and should indeed be decoupled from the efficiency indicator by correcting the energy trends, as also done in this paper. 2.5. Data availability Finally, despite the existence of methods and tools for energy accounting, information on the characteristics of the building stock and its energy use is lacking, and data broken down by subsectors, types and end-uses are rarely available and sometimes unreliable. Therefore, data availability is a determinant factor for the choices and methodological decisions previously described. In fact, finding information on the building sector from existing sources remains a challenge, resulting in few cross-country studies for this sector compared to industry and transport [16]. This paper contributes to filling this gap by providing a cross-country analysis of the regions with the highest consumption in the world (China, the US and the EU). 3. Method and materials 3.1. Pyramidal approach A pyramid of indicators is proposed to progressively decompose the energy consumption of buildings (E B ) (Fig. 1). For each level, the changes in the energy use (ΔE B ) are disaggregated to analyse the effect of different drivers. This approach differs from others in the literature by proposing a stepwise methodology that allows for a better and more comprehensive understanding of the driving forces, in contrast with the usual practice of stand-alone decompositions. To quantify the drivers’ contributions, the LMDI I [12] is employed due to its desirable properties, including perfect decomposition (with no residual term), consistency in aggregation, numerical robustness, and ease of use, which have established it as a standard in the energy statistics research field. At the top of the pyramid, the simplest decomposition consists of the disaggregation of buildings’ energy consumption into the effects of the main activity and efficiency factors. Although energy is used in buildings to provide different services (comfort, lighting, hot water, cooking, etc.), the floor area is chosen as the main activity indicator. The choice is supported by the fact that it is the main driver for the demand of Heating, Ventilation and Air Conditioning (HVAC), which represented 38 % of global buildings’ consumption in 2020 [44] and therefore the highest consuming end-use. Consequently, it is used to define the sectoral efficiency indicator: the energy-use intensity. This way, the less energy the sector uses to provide the building services demanded in a given space, the more efficient it is. EB=A⋅EB A=A⋅eB(3) where E B is buildings’ energy consumption, A is floor area and e B [kWh/m 2 ] is the energy intensity of buildings (energy-use intensity). Then, changes in buildings’ energy consumption according to the first level of the pyramid can be decomposed into the floor area effect (ΔEBA) and the efficiency effect (ΔEBeB) (Eq. (4)), where their contributions can be assessed by applying the LMDI as in Eqs. (5) and (6). ΔEB=ΔEBA+ΔEBeB(4) ΔEBA=L(Et B,E0 B)⋅ln(At A0)(5) ΔEBeB=L(Et B,E0 B)⋅ln(et B e0 B)(6) with L(a, b) the log mean difference between a and b (Eq. (7)), E t B and E 0 B the buildings’ energy consumption at time t and 0, A t and A 0 the floor area at time t and 0, and e B t and e B 0 the energy-use intensity at time t and 0, respectively. L(a,b) = a−b ln(a/b)(7) At the second level, the floor area is decomposed into the effects of population and urbanisation (u), defined as the area per capita [m 2 /cap]. EB=P⋅A P⋅EB A=P⋅u⋅eB(8) This allows the change in the energy use to be broken down into three factors, disaggregating the activity drivers (lifestyle factors) (Eq. (9)). In addition to the impact of the energy intensity (calculated as in Eq. (6)), the contribution of the population (ΔEBP) and urbanisation (ΔEBu) can be assessed by applying the LMDI as in Eqs. (10) and (11). ΔEB=ΔEBP+ΔEBu+ΔEBeB(9) ΔEBP=L(Et B,E0 B)⋅ln(Pt P0)(10) ΔEBu=L(Et B,E0 B)⋅ln(ut u0)(11) where P t and P 0 are the population at time t and 0, and u t and u 0 are the urbanisation (area per capita) at time t and 0, respectively. At the third level, the number of buildings is introduced to split urbanisation into the effects of the number of buildings per capita, b [build/cap] (the inverse of occupancy), and the buildings size, a [m 2 / build]. Thus, urbanisation may grow due to an increase in the buildings size or a reduction of the occupancy (more buildings per capita). Fig. 1. Pyramid approach for the decomposition of the energy consumption in buildings. M. Gonz´ alez-Torres et al. Energy & Buildings 332 (2025) 115461 5 EB=P⋅B P⋅A B⋅EB A=P⋅b⋅a⋅eB(12) The decomposition of energy consumption changes at this stage adds then the effects of the buildings per capita (ΔEBb) and buildings size (ΔEBa) (Eq. (13)). To evaluate their contributions, the LMDI can be applied as shown in Eqs. (14) and (15), while the impact of the energy intensity and population can be assessed as in previous Eqs. (6) and (10). ΔEB=ΔEBP+ΔEBb+ΔEBa+ΔEBeB(13) ΔEBb=L(Et B,E0 B)⋅ln(bt b0)(14) ΔEBa=L(Et B,E0 B)⋅ln(at a0)(15) where b t and b 0 are the buildings per capita at time t and 0, and a t and a 0 are the buildings size at time t and 0, respectively. Then, the effect of climate (c) is introduced, by correcting and normalising the energy consumption (E Bc ) assuming a linear regression with heating degree days (HDD) [45], and thus the energy-use intensity (e Bc ). EB=P⋅B P⋅A B⋅EBc A⋅EB EBc =P⋅b⋅a⋅eBc⋅c(16) The decomposition of energy consumption changes at the fourth level is presented in Eq. (17), including the effect of the corrected efficiency (ΔEBeBc ) and climate (ΔEBc). Their contributions can be computed by application of the LMDI as in Eqs. (18) and (19), while the impact of the population, buildings per capita and buildings size can be calculated as in previous Eqs. (10), (14) and (15), respectively. ΔEB=ΔEBP+ΔEBb+ΔEBa+ΔEBeBc +ΔEBc(17) ΔEBeBc =L(Et B,E0 B)⋅ln(eBct eBc0)(18) ΔEBc=L(Et B,E0 B)⋅ln(ct c0)(19) where e B t and e B 0 are the weather-corrected energy-use intensities (corrected buildings’ energy use per area) at time t and 0 and c t and c 0 are the climate indicator (as the ratio of the buildings energy use to that corrected assuming a linear regression with heating degree days) at time t and 0, respectively. Finally, the effect of the structural changes among residential and tertiary subsectors is evaluated: EB=P⋅B P⋅A B⋅∑ s As A⋅EBcs As ⋅EBs EBcs =P⋅b⋅a⋅∑ s ss⋅eBcs⋅cs(20) where A s is the area of building subsector s (tertiary or residential), E Bcs is its weather-corrected energy consumption, E Bs is its actual energy consumption, s s is its share in the total floor area, e Bcs is its corrected energy-use intensity, and c s is its climatic factor. Thus, the contributions of the drivers to consumption changes are assessed as follows, allowing for the quantification of the effect of population (ΔEBPʹ), occupancy (ΔEBbʹ), buildings size (ΔEBaʹ), structure (ΔEBs), efficiency (ΔEBeBcʹ) and climate (ΔEBcʹ). ΔEB=ΔEBPʹ+ΔEBbʹ+ΔEBaʹ+ΔEBs+ΔEBeBcʹ+ΔEBcʹ(21) ΔEBPʹ=∑ s L(Et Bs,E0 Bs)⋅ln(Pt P0)(22) ΔEBbʹ=∑ s L(Et Bs,E0 Bs)⋅ln(bt b0)(23) ΔEBaʹ=∑ s L(Et Bs,E0 Bs)⋅ln(at a0)(24) ΔEBs=∑ s L(Et Bs,E0 Bs)⋅ln(st s s0 s)(25) ΔEBeBcʹ=∑ s L(Et Bs,E0 Bs)⋅ln(et Bcs e0 Bcs)(26) ΔEBcʹ=∑ s L(Et Bs,E0 Bs)⋅ln(ct s c0 s)(27) where E t Bs and E 0 Bs are the buildings’ energy consumption of subsector s at time t and 0; P t and P 0 are the population at time t and 0; b t and b 0 are the buildings per capita at time t and 0; a t and a 0 are the buildings size at time t and 0, s s t and s s 0 are the shares of the floor area of subsector s at time t and 0; e Bcs t and e Bcs 0 are the corrected energy-use intensities of subsector s at time t and 0; and c s t and c s 0 are the climatic factors of subsector s at time t and 0. Note that, in order to facilitate the comparative analysis across the different regions in the results section, the relative changes and the percentage contributions of each driver are calculated. For this purpose, the figures are divided by the initial value of the aggregate variable E B 0 . 3.2. Data Collecting energy and activity information for buildings from the existing data sources is a major challenge. The indicators needed are difficult to measure, so the data are hard to find or unavailable, especially in developing countries [41]. The main limitations concern information on the building stock, such as the number of buildings or the floor area. The difficulties are especially critical in the tertiary subsector, due to the difficulties in gathering data on buildings that are multi-tenanted and host different activities. Data collection in this field needs to be improved, as the development, implementation and monitoring of policies to limit energy consumption growth can only succeed if they are based on the relevant disaggregated analysis of the information. The geographical coverage of this paper is highly subject to data availability. Consequently, the proposed methodology can only be applied to China, the US and the EU and for the different time periods for which the required information was found. Table 2 presents the main sources used and the time periods available for each of the indicators for which raw data are needed. Note that the information for the whole building sector always appears broken down into ‘residential’ and ‘tertiary’ and the data need then to be aggregated. The rest of the indicators in the methodology can be calculated as ratios according to their definitions in Table 1. For the US, the main data sources are the Residential Energy Consumption Survey (RECS) [46] and the Commercial Buildings Energy Consumption Survey (CBECS) [47], conducted by the U.S. Energy Information Administration (EIA). They collect highly detailed and disaggregated data on buildings’ characteristics, consumption and expenditure that, unfortunately, cannot be released on a yearly basis due to their preparation, collection and processing time and cost. Based on those and complemented by information gathered through targeted questionnaires, the IEA produces the estimated trends used in this work, in the Energy End-uses and Efficiency Indicators database [48]. In the EU, Enerdata is a private company that releases yearly activity and energy data in the Odyssee database [49], funded by the European Commission’s Horizon 2020 programme. However, it lacks some relevant information for this study, such as the number of tertiary buildings. More detailed and disaggregated information is being elaborated by the EU Building Stock Observatory (BSO) [50] managed by the European Commission’s Directorate-General for Energy. Unfortunately, at the M. Gonz´ alez-Torres et al. Energy & Buildings 332 (2025) 115461 6 time of writing, the data from these sources cannot be combined to complement each other, owing to discrepancies for instance due to the blurred line between permanently/temporarily occupied and vacant dwellings. Moreover, while the BSO already contains data on the number of tertiary buildings for 2020, it still does not provide trends, thus being insufficient for the decomposition proposed in this paper. Nevertheless, the database will be upgraded regularly to expand the data and improve the user experience, hopefully addressing the identified gaps. In China, there are not official statistics and databases on building energy use for historical reasons, and the building sector does not appear in the national energy statistics and balance sheets [51]. However, the Building Energy Research Center (BERC) of Tsinghua University has been working on data of the building sector since 2005, publishing annual reports also in English since 2016. Among these publications, the following have served to gather the data for the decompositions in this study: Jiang et al. (2018) [52], Hu et al. (2022) [51], and the China Building Energy and Emission Yearbook 2023 [17]. Unfortunately, these also lack information on the number of tertiary buildings, since the floor area is used as the only activity indicator for the subsector. In addition, other international sources are used, namely the World Development Indicators database [53] by the World Bank for the population, the Weather for Energy Tracker [54] by the IEA and Fondazione Euro-Mediterraneo Sui Cambiamenti Climatici (CMCC) for the Heating Degree Days, and the IEA World Energy Balances [55] for residential and tertiary energy consumption. Based on the available data, the decompositions are performed for the following the time periods: 2003 – 2018 in the US, 2000 – 2022 in the EU and 2000 – 2021 in China. 4. Results Global buildings’ energy use is strongly influenced by the trends in China, the US and the EU, which are the highest consumers, accounting for 45 % of the consumption in 2021 (Fig. 2). In China (20 EJ in 2021), buildings’ energy use rose sharply after its economic expansion and industrialisation, increasing by 58 % since 2000 and surpassing the figures in the US (19.9 EJ) and the EU (16.3 EJ). In contrast, consumption in the US and the EU only grew by 4 % and 1.9 %, respectively, over the period under study. As a direct response to economic development, the buildings’ floor area expanded in every region under study, as shown by the decomposition in the first level of the pyramid approach (Fig. 3). China shows impressive growth in the floor area (114 %) due to the increasing wealth and the migration from rural to urban locations, which led to lifestyle changes and an increase in personal living space [52]. However, the higher standards did not immediately translate into a higher demand for energy services, resulting in a decreasing energy-use intensity due to a faster growth in the area than in the consumption. Consequently, the contribution of the floor area (97 %) was partially offset by the effect of the energy-use intensity (−39 %), limiting the Chinese consumption Table 2 Data sources and time periods available by indicator. Indicator United States European Union China Source Time period Source Time period Source Time period Population World Bank −World development indicators (2024) [53] 1960–2023 World Bank −World development indicators (2024) [53] 1960–2023 World Bank −World development indicators (2024) [53] 1960–2023 Floor area Residential IEA −Energy End-uses and Efficiency Indicators (2023) [48] 2000–2021 Enerdata −Odyssee (2024)* [49] 1990–2022 Jiang et al. 2018 [52] 2000 Hu et al. 2022 [51] 2010 – 2020 China Building Energy and Emission Yearbook (CBEEY) (2023) [17] 2021 Tertiary IEA −Energy End-uses and Efficiency Indicators (2023) [48] 2000–2021 Enerdata −Odyssee (2024) [49] 2000–2022 Jiang et al. 2018 [52] 2000 Hu et al. 2022 [51] 2010 – 2020 EIA −Commercial Buildings Energy Consumption Survey (CBECS) [47,56,57] […] 2003, 2012, 2018 CBEEY (2023) [17] 2021 Number of buildings Residential IEA −Energy End-uses and Efficiency Indicators (2023)** [48] 2000–2021 Enerdata −Odyssee (2024) *** [49] 1990–2022 Jiang et al. 2018 [52] 2000 Tertiary EIA – CBECS [47,56,57] […] 2003, 2012, 2018 EU Building stock observatory [50] 2020 Hu et al. 2022 [51] 2010–2020 Heating Degree Days IEA −Energy End-uses and Efficiency Indicators (2023) [48] 2000–2021 Enerdata −Odyssee (2024) [49] 1990–2022 IEA, CMCC −Weather for energy tracker (2024) [54] 2000–2024 Energy consumption Residential IEA −World Energy Balances (2023) [55] 1971–2021 Enerdata −Odyssee (2024) [49] 1990–2022 IEA −World Energy Balances (2023) [55] 1971–2021 Tertiary IEA −World Energy Balances (2023) [55] 1971–2021 Enerdata −Odyssee (2024) [49] 1990–2022 IEA −World Energy Balances (2023) [55] 1971–2021 Methodological notes: * average, then multiplied by the number of dwellings to obtain the total, ** occupied dwellings, *** permanently occupied dwellings, **** number of households. Fig. 2. Trends of final energy consumption in buildings in the US, China, the EU and the rest of the world (2000–2021). Sources: Odyssee [49], IEA [55]. M. Gonz´ alez-Torres et al. Energy & Buildings 332 (2025) 115461 7 growth to 58 % (7.3 EJ) from 2000 to 2021. In the US and the EU, the construction boom prior to the period under study kept their floor area increase below that in China (9 % and 27 %, respectively), contributing to rising consumption by 9 % and 24.2 %. In addition to the floor area growth, wealth enabled the spread of efficient but expensive equipment and building designs. It resulted in energy savings at no cost to the welfare of buildings’ occupants [58], which contributed to counteracting the effect of the floor area growth by −5% and –22.3 %, respectively, limiting the increase in consumption to 4 % (0.76 EJ) and 1.9 % (0.28 EJ). Fig. 4 shows the relationship between trends of national energy-use intensity and floor areas, which would lead to constant lines of energy consumption in buildings if they evolved at the same rate. The increases in the buildings’ floor areas shown in the previous figure are associated with mainly upward tendencies in China and the EU, while fluctuations are observed in the US. In absolute figures, the energy-use intensity in developed regions such as the EU and the US (around 180 kWh/m 2 ) contrasts with that of China (82 kWh/m 2 ), due to energy conservation habits rather than to higher efficiency levels [59,60]. Larger floor area figures in China (3.5 times those of the EU and 6.5 times those of the US) correspond to a larger population, leading to lower per capita figures. Alarmingly, if the current Chinese building stock adopted energy-use intensity levels equivalent to those in the US and the EU as living standards tend to converge, it would result in more than twice the energy consumption, over 12 PWh (43.2 EJ). At the second level of the pyramid (Fig. 5), the area growth is decomposed to analyse the effects of population and urbanisation. The population increased by 13 % in the US, 4 % in the EU and 12 % in China. This demanded higher consumption levels and contributed to rising the energy use in every region, namely by 12 % in the US, 4.2 % in the EU and 14 % in China. Urbanisation, coupled with the effect of population in the EU and China, contributed to increasing consumption by 20 % and 83 %, respectively. In contrast, it decreased in the US, contributing −3% of the change in consumption, so slightly compensating the population effect. The impressive growth of urbanisation in China is explained by the particularly low figures in 2000 (25 m 2 /capita, half the European and a quarter of the American), which almost doubled to 48 m 2 /capita in the period under study. This allowed the narrowing of the differences to 10 % with the EU and 50 % with the US, reducing inequality and demanding for larger dwellings and more spaces for education, health, leisure, etc. (Fig. 6). As for the annual trends, China consistently maintained an uninterrupted growth throughout the entire period under consideration. Meanwhile, the decreasing urbanisation in the US over the period decomposed saw two turning points. The upward trend registered until 2005 was balanced out after the Great Recession that affected the housing market in order to adapt to the challenging financial conditions. However, the trend shows an increase again, especially since 2014, suggesting further growth and consequent increases in energy consumption. Conversely, the EU shows a period of growth at the beginning of the study period, which marked the increase depicted in Fig. 5, whereas it is stable from 2012, suggesting a future constant tendency. Following the progressive decomposition, the effect of urbanisation is disaggregated into that of the buildings’ occupancy and size (Fig. 7). However, this step can be only conducted for the US, due to the lack of data for the other regions. The drop in urbanisation in the US (−3%) was caused by the reduction of building sizes (contributing −4%), which was partially Fig. 3. Decomposition of changes in buildings’ energy consumption (E B ) into floor area (A) and energy-use intensity (e B ) in the US, the EU and China. Based on CBEEY [17], IEA [48,55], Odyssee [49] and Jiang et al. 2018 [52] data. Fig. 4. Energy-use intensity vs. buildings floor area in the US, the EU and China. Based on CBEEY [17], IEA [48,55], Odyssee [49], Hu et al. 2022 [51] and Jiang et al. 2018 [52] data. Fig. 5. Decomposition of changes in buildings’ energy consumption (E B ) into population (P), urbanisation (u) and energy-use intensity (e B ) in the US, the EU and China. Based on CBEEY [17], IEA [48,55], Odyssee [49], Jiang et al. 2018 [52] and World Bank [53] data. Fig. 6. Urbanisation trends in the US, the EU and China. Based on CBEEY [17], IEA [48], Odyssee [49], Hu et al. 2022 [51], Jiang et al. 2018 [52] and World Bank [53] data. M. Gonz´ alez-Torres et al. Energy & Buildings 332 (2025) 115461 8 compensated by lower occupancy rates (contributing 1 %), i.e. more buildings per capita. Less occupied buildings consume more by not sharing energy services and equipment (mainly HVAC) [61]. As this decomposition cannot be performed for China and the EU, the diverse dynamics of the building stock in these regions cannot be analysed, making it impossible to define adaptive strategies to cope with the challenges posed by their particular contexts and patterns. Fig. 8 illustrates the fourth level of the pyramid approach, in which the energy-use intensity is corrected to isolate the climatic effect on consumption changes. In the US, a slightly colder year in 2018 compared to 2003 contributed 1 % of the increase in consumption, although this effect was successfully compensated by the drop in its corrected intensity (−6%). In contrast, the decrease in the energy-use intensity in the EU (–22.3 %) was a result of the warmer weather (i.e. lower number of HDD) and of the significant improvement of the normalised intensity, which contributed −1.9 % and −20.4 % of the change in the energy-use intensity, respectively. However, the correction of the energy consumption to neutralise the effects of the weather by assuming a linear regression with heating degree days does not always work [62]. It can even lead to unrealistic fluctuations in developing countries, where the response to weather variations does not necessarily translate into increased energy use, but rather into decreased thermal comfort, as low income levels restrict energy expenditure. For this reason, the results of this decomposition in China are not meaningful and should be neglected. Although the HDD decreased between 2000 and 2021, climate appeared to have significantly increased consumption (by 29 %), compensated by the improvement of the normalised intensity (−68 %). Finally, the structure of the building stock is also analysed as a driver of the energy consumption. Higher shares of the most intensive building types would increase sectoral energy consumption. For instance, tertiary buildings in the US are twice as intensive (296 kWh/m 2 ) as residential ones (137 kWh/m 2 ), while the sectoral differences are less significant in China and the EU (Table 3). The particularly low energy-use intensity of Chinese tertiary buildings, even below that of residential ones, can be explained by the explosion in demand for services that accompanied the country’s rapid economic development, which resulted in a construction boom of poorly equipped buildings. Consequently, their energy consumption is around three times lower than it would be if they had the levels of comfort and equipment they have in the US or the EU. In the future, the continued increase in wealth will allow them to benefit from more energy services and will increase their energy-use intensity to approach that of developed regions. Consequently, the decomposition according to the last level of the pyramid (Fig. 9) shows that the shift towards tertiary buildings in the US (from 23 % to 27 % of the building stock) drove a 4 % rise in consumption, which was compensated by the efficiency gains. In the EU, the slight drop in the tertiary floor area (from 27.4 % to 26.8 %) resulted in a minor structural effect that offset the energy use increase by −0.2 %. Finally, the significant growth of the tertiary share in China (from 14 % to 22 %) compensated the consumption growth by −3%, due to the lower tertiary energy-use intensities. Therefore, the corrected energyuse intensities, once the weather and structural effects have been discounted, are shown to have contributed to reduced energy consumption in all regions, namely by 10 % in the US, 20.2 % in the EU and 35 % in China, providing a better indication of the evolution of the efficiency in the sector. 5. Discussion and policy implications Over the past 40 years, the US energy policy for buildings has put in place numerous measures to promote the energy efficiency of buildings and building equipment. Since 2000, these policies have led to significant reductions in energy-use intensity (to 180 kWh/m 2 ), but the degree of urbanisation remains very high (~100 m 2 /cap) (both in residential and non-residential buildings). Consequently, the US remains the largest per capita consumer in buildings (~17 400 kWh/cap), and the situation is aggravated by the population growth, which has increased by 18 % over the last 25 years. The results suggest that it will be difficult to reduce the impact of the building sector in the US without changing the American way of life, since it would require relying on almost technologically unattainable efficiency improvements. It is therefore essential to reduce the energy demand through sufficiency measures and the use of clean energy sources. Although sufficiency is gaining momentum around the world, specific measures are still relatively incipient in the US building sector, and in any case voluntary. Among the practices aligned with sufficiency principles, the installation of demand-side energy management systems could be incentivised while passive and compact design of buildings Fig. 7. Decomposition of changes in buildings’ energy consumption (E B ) into population (P), number of buildings per capita (b), buildings size (a) and energy-use intensity (e B ) in the US. NB. Data limitations prevent this disaggregation level for the EU and China. Based on IEA [48,55], EIA [47,56] and World Bank [53] data. Fig. 8. Decomposition of changes in buildings’ energy consumption (E B ) into population (P), number of buildings per capita (b), buildings size (a), weathercorrected energy-use intensity (e Bc ) and climate (c) in the US. NB. Data limitations only allow consumption in the EU and China to be decomposed into population (P), urbanisation (u), weather-corrected energy-use intensity (e Bc ) and climate (c). Based on CBEEY [17], IEA [48,55], Odyssee [49], Jiang et al. 2018 [52], EIA [47,56], World Bank [53] and IEA and CMCC [54] data. Table 3 Energy-use intensity of residential and tertiary buildings. Residential (kWh/m 2 ) Tertiary (kWh/m 2 ) US (2018) 137 296 EU (2022) 162 221 CHINA (2021) 83 76 Based on CBEEY [17], IEA [48,55], Odyssee [49], EIA [47,56], IEA [48,55] data. M. Gonz´ alez-Torres et al. Energy & Buildings 332 (2025) 115461 9 could be promoted by certification schemes such as the US Leadership in Energy and Environmental Design (LEED). Other solutions include the reuse and retrofit of existing buildings, which would curb the urbanisation trend, and behavioural changes, such as adjusting thermostats, turning off lights and reducing demand for domestic hot water. Lately, the focus has moved to also reducing emissions and promoting renewable energy, with an increasing emphasis on a broader concept of sustainability and smart technologies. While there are several measures to encourage on-site renewables in US buildings (e.g. tax credits and incentives for the installation of solar panels or government programmes to support research and development in renewable technologies), there is still room for further expansion and innovation. There is no specific federal mandate requiring new constructions to be net-zero energy or emissions buildings, missing the opportunity to accelerate the transition in that direction. In the EU, policies to reduce the energy consumption of buildings have been implemented since the 1990 s [63]. The Energy Performance of Buildings Directive (EPBD) 2 sets minimum renewable shares and energy performance standards, encouraging energy-efficient design and renovation. The Energy Efficiency Directive (EED) 3 establishes a framework for improving energy efficiency in various sectors, with specific targets for public buildings. The Renewable Energy Directive (RED) 4 promotes the use of renewable energy sources, including sectorspecific targets in buildings. In addition, the EU offers various financial instruments, such as loans and grants, to support energy efficiency projects in buildings and the EU Taxonomy 5 contributes to guarantee the environmental benefit of green investments. Since 2000, these policies, combined with national and regional initiatives, have contributed to a reduction in energy-use intensity down to ~ 180 kWh/m 2 . Urbanisation in the EU has grown steadily this century to about 55 m 2 /capita, but is still about half that of the US. Its growth, coupled with a small increase in population, increased the energy use in buildings, but has been offset by energy efficiency improvements to reduce consumption since 2009. European citizens now consume 9 400 kWh/capita, and new policies are obviously needed to maintain and accelerate the trend. The pyramidal decomposition approach points to urbanisation as the main factor driving up the energy consumption through the period studied. As 85 % of European buildings were built before 2000 and, of them, 75 % have a poor energy performance, policymakers should focus on renovation of existing buildings over new construction. Parts of these issues have been addressed with the adoption in May 2024 of the new Energy Performance Building Directive 6 (EPBD recast). Besides the main targets of a zero-emission and fully decarbonised building stock by 2050, there are objectives that could enhance actions towards these targets. Among them, the Directive aims to create a stable environment for solid investment decisions and the introduction of a building renovation passport, crucial actions to support deep renovation strategies, both in residential and non-residential buildings. In parallel, these efforts should be coupled with the installation of more efficient equipment, which needs to be promoted through minimum performance and information requirements [64]. Contributing to this purpose, in July 2024 the Ecodesign for Sustainable Products Regulation (ESPR) 7 entered into force, aiming to be the cornerstone of the Commission’s approach to more environmentally sustainable and circular products. In addition, it is important to highlight that the analysis and decomposition of energy consumption’s trends are meaningful exercises that the European Commission carries out on a yearly basis. The LMDI results are regularly used to inform policy Directorates-General on the drivers of energy consumption trends towards 2030 and 2050 targets and to disseminate information through the State of the Energy Union report. In conclusion, the policy intentions in the EU seem to be going in the right direction, while their continuity and correct implementation need to be ensured. A significant challenge stems from the fact that the EU comprises 27 Member States with their own national laws. While EU directives need to be transposed into Member States’ national laws, discrepancies in ambition and implementation levels persist. For instance, Maduta et al. [65] reported how the energy performance levels of Nearly-Zero Energy Buildings (NZEBs) defined according to the EPBD vary widely across the EU countries. Although the Directive allows for the definition of flexible thresholds to take into account national, regional or local climate, social and economic conditions, many countries lag behind in meeting recommended benchmarks. To address this issue, the European Commission should enhance the monitoring and harmonisation of the national transpositions of the Directives as well as apply effective, proportionate and dissuasive penalties, since targeted national revisions could drive significant improvements in energy performance in buildings. As for China, energy efficiency policies so far have been very effective (from 110 kWh /m 2 to 80 kWh/m 2 ) but have not been able to compensate for the floor area growth. Such policies include energy efficiency standards for new buildings and 5-year plans to set national targets for reducing energy intensity and promoting energy-efficient technologies. Beside this, the low energy intensity figures, approximately half those of developed countries, can be explained by the sociocultural habits of the Chinese population and economic aspects, which result in a rational use of energy (3 900 kWh/cap). In fact, the results of the pyramidal approach point to the increased floor area as the main driver of energy consumption in Chinese Fig. 9. Decomposition of changes in buildings’ energy consumption (E B ) into population (P), number of buildings per capita (b), building size (a), structure (s), weather-corrected energy-use intensity (e Bc ) and climate (c) in the US. NB. Data limitations only allow consumption in the EU to be decomposed into population (P), urbanisation (u), structure (s), weather-corrected energy-use intensity (e Bc ) and climate (c), while Chinese decomposition is limited to population (P), urbanisation (u), structure (s) and energy-use intensity (e B ). Based on CBEEY [17], IEA [48,55], Odyssee [49], Jiang et al. 2018 [52], EIA [47,56], World Bank [53] and IEA and CMCC [54] data. 2 Directive 2010/31/EU of the European Parliament and of the Council of 19 May 2010 on the energy performance of buildings (recast). 3 Directive (EU) 2023/1791 of the European Parliament and of the Council of 13 September 2023 on energy efficiency and amending Regulation (EU) 2023/ 955 (recast). 4 Directive (EU) 2023/2413 of the European Parliament and of the Council of 18 October 2023 amending Directive (EU) 2018/2001, Regulation (EU) 2018/ 1999 and Directive 98/70/EC as regards the promotion of energy from renewable sources, and repealing Council Directive (EU) 2015/652. 5 Regulation (EU) 2020/852 of the European Parliament and of the Council of 18 June 2020 on the establishment of a framework to facilitate sustainable investment, and amending Regulation (EU) 2019/2088. 6 Directive (EU) 2024/1275 of the European Parliament and of the Council of 24 April 2024 on the energy performance of buildings (recast). 7 Regulation (EU) 2024/1781 of the European Parliament and of the Council of 13 June 2024 establishing a framework for the setting of ecodesign requirements for sustainable products, amending Directive (EU) 2020/1828 and Regulation (EU) 2023/1542 and repealing Directive 2009/125/EC. M. Gonz´ alez-Torres et al.