Analysis of the EU Residential Energy Consumption: Trends and Determinants

This article analyses the status and trends of the European Union (EU) residential energy consumption in light of the energy consumption targets set by the EU 2020 and 2030 energy and climate strategies. It assesses the energy efficiency progress from 2000 to 2016, using the official Eurostat data. In 2016, the residential energy consumption amounted to 25.71% of the EU’s final energy consumption, representing the second largest consuming sector after transport. Consumption-related data are discussed together with data on some main energy efficiency policies and energy consumption determinants, such as economic and population growth, weather conditions, and household and building characteristics. Indicators are identified to show the impact of specific determinants on energy consumption and a new indicator is proposed, drawing a closer link between energy trends and policy and technological changes in the sector. The analysis of these determinants highlights the complex dynamics behind the demand of energy in the residential sector. Decomposition analysis is carried out using the Logarithmic Mean Divisia Index technique to provide a more complete picture of the impact of various determinants (population, wealth, intensity, and weather) on the latest EU-28 residential energy consumption trends. The article provides a better understanding of the EU residential energy consumption, its drivers, the impact of current policies, and recommendations on future policies.


Introduction and Policy Background
Global residential energy demand has steadily increased over the past decades. [1,2]. A generalized higher level of basic comfort in the last years has translated to increased consumption in the sector. Larger homes, new services, and new and larger appliances and equipment, including the direct rebound effect, have boosted the demand for energy [3]. At the same time, European countries experienced an increase in the energy efficiency of residential appliances, ICT (Information, Communication and Technology), lighting, and heating and cooling equipment mainly due to the implementation of energy efficiency policies [4]. Technical progress on efficiency, changes in consumer behaviour, socio-economic characteristics, and policy interventions play an interconnected and key role in reducing residential energy consumption growth and in achieving energy savings.
Looking at the policy dimension, we observe a growing number of policy commitments and strategies, calling for action at local, regional, national, and global levels. Several policies and measures , dwelling-related characteristics (Figures 11 and 12), and climatic conditions (Figure 13), are identified and analysed on a yearly basis for each member state. Similarities in trends between the residential FEC and specific determinants are drawn to provide evidence on the correlation between variables. Consequently, these potential correlations are analysed using the regression analysis in an attempt to understand which determinants have the most pronounced impact on the residential FEC trends (Figures 14-17). Linear regression correlation diagrams are applied for a specific year comparing the EU-28 member states. Finally, the residential FEC indicator proposed by the authors is analysed alongside all the described determinants (Figures 18 and 19), with the aim to assess how energy consumption trends change if all the potential determinants are considered. If this indicator shows a declining trend in the time period studied, it could be viewed as evidence that the EU's energy efficiency policies implemented in previous years, in addition to technological developments, have had a considerable impact. Decomposition analysis is also applied to support the results ( Figure 20).

Impact of Policies on Energy Consumption
A limited number of articles have assessed the impact of EU policies on residential energy consumption. Bertoldi and Hirl [2], Bertoldi and Mosconi [21], and Horowitz and Bertoldi [22] used panel data to identify the energy savings resulting from EU policies, using a counterfactual line and an energy efficiency policy indicator based on the Odex index as an explicit variable or on the adopted policies as reported in the MURE (Mesures d'Utilisation Rationnelle de l'Energie) database. Filippini and Hunt [23] estimated the efficiency of US residential energy consumption by using an econometric energy demand model, which includes, beside other determinants, the "underlining energy efficiency". This was estimated by adopting a stochastic frontier function. Filippini also concluded that energy intensity and energy per capita are not good indicators. In another article, Fillippini et al. [24] assessed the level of energy efficiency of the EU residential sector against the potential for energy savings by using stochastic frontier analysis. The authors also estimated the impact of energy efficiency policy. The authors concluded that financial incentives are important for energy efficiency investments together with energy performance standards, while information measures do not have a significant impact. Laes et al. in their article [25] reviewed the effectiveness of individual policies or policy packages for CO 2 emission reduction and/or energy savings. Based on the literature review carried out in the paper, the authors concluded that financial incentives and subsidies have a positive impact on energy efficiency improvements. Energy and CO 2 taxation is another important measure to induce energy savings and reduce CO 2 emissions, since the reviewed studies indicate that people are responsive to savings in energy costs and expected price increases in the future. The authors concluded that there is no quantitative evidence on the impact of information measures.
Other authors assessing the impact of specific policies include Aydin and Brounen [26]. Aydin and Brounen investigated the electricity and non-electricity energy consumption separately and focused on two distinct types of regulatory measures: Mandatory energy efficiency labels for household appliances and building standards. The authors concluded that both appliance energy labelling and stringent building standards reduce residential energy consumption. Ó Broin et al. [27] analysed space heating energy efficiency policies in the EU and member states in the period of 1990-2010 using a panel of 14 EU countries. Policies were subdivided into financial, regulatory, and informative policies. The authors showed that the impact of regulatory policies is stronger than the impact of financial policies and information measures.

Impact of Drivers on Energy Consumption
A number of authors assessed the impact of specific drivers on residential energy consumption. Belaid [28] assessed the direct and indirect determinants of the residential energy consumption in France by using a structural equation modelling approach to identify the impact of dwelling characteristics, household attributes, climate, and behaviour on residential energy use. He concluded that the direct effect of household-related attributes on energy consumption is lower than the effect of dwelling attributes. Román-Collado and Colinet [29] investigated the impact of productivity and residential living standards on energy consumption by using the Logarithmic Mean Divisia Index (LMDI-I) method, and concluded that the residential standard of living, i.e., the income per capita, is a driver for residential energy consumption. Borozan [30] investigated the determinant of residential energy consumption. The analysis confirmed that the socio-economic (e.g., disposable income, etc.) and contextual variables (e.g., climatic conditions) are important determinants of energy consumption. Brounen et al. [31] analysed residential gas and electricity determinants. The authors concluded that residential gas consumption is mainly determined by the physical characteristics of the building, while electricity consumption is more related to household composition, in particular income and family composition. Lévy and Belaid [32] analysed the impact on energy consumption of household profiles and building profiles by investigating global consumption, consumption per m 2 , and per capita. The study highlighted the impact of the demographic characteristics of households. Filippini and Hunt's residential energy demand model [23] identified the following drivers: Income, price, population, average household size, heating degree days, cooling degree days, and the share of detached houses. Otsuka [33] estimated the Japanese residential energy demand and energy efficiency levels using a stochastic frontier function and investigated the regional determinants by controlling for income, price, household size, climate, and urbanization. In addition, Otsuka identified the population concentration and electrification rates as drivers for residential energy efficiency. Finally, Reuter et al. [34] showed that comfort and behavioural effects in households in the EU, linked to the size of dwellings and persons per dwelling, drove up energy consumption, while improved efficiency, i.e., lower energy consumption for space heating per unit of floor area, had a major inhibiting effect.

Methodology
This article studies the EU residential energy consumption trends and possible determinants (economic conditions, population status, climatic conditions, household characteristics, energy prices). This is a fundamental step in evaluating if EU and national energy policies have had a positive impact on energy consumption trends in the context of the EU 2020 energy strategy. The article focuses on the period from 2000 to 2016; the year, 2016, is the latest year for which statistics were available when the analysis was carried out. For the completion of the article, an important amount of data was selected, elaborated, and analysed. The data used for the identification of the main determinants of the residential energy consumptions were extracted from two databases: The Eurostat Database [35] and the Odyssee database [36,37]. The Eurostat Database is available online and provides high quality data and statistics about EU member states, enabling a direct comparison between them. In particular, the Eurostat Energy Balances, which are updated in January and June every year, were used to extract a variety of energy data. The update of the database in June 2018 was used. The database of the Odyssee project, on the other hand, is managed by Enerdata, and contains detailed energy efficiency and energy indicators and data on energy consumption, activity indicators, and their related CO 2 emissions. The Odyssee database includes, among others, data taken from national sources and from Eurostat. The Odyssee energy data used in this article were extracted from the Odyssee database of October 2018.
Both the Eurostat and Odyssee databases provide energy data for the main economic sectors (residential/households, services, industry, and transport) and for the EU-28 member states over the period of 1990-2016. This enables the comparison of the energy consumption trends between member states and between economic sectors. Eurostat was the main data source used in this article; however, Odyssee datasets were used as a complementary source, especially in cases of more elaborate energy indicators or detailed data, which are not currently covered by Eurostat. The article starts from the analysis of the primary and final energy consumption trends at the EU level. These are compared with the targets defined under the EED. In this context, it should be noted that there is a mismatch between the official EU28 targets (as defined in 2013 with the accession of Croatia) and the sum of the indicative targets communicated and updated by member states. The sum of indicative member states' primary energy targets is 3% greater than the EU28 target, while it is the opposite (−1%) for final energy targets [38]. To reply to the question, 'Is the EU on track to achieve its objectives?', the linear regression path between the historical data of 2005 (the year before the transposition of the EPBD) and the 2020 target was considered. This linear regression path offers a theoretical "first-order" approximation path to the targets.
The analysis focuses on two significant indicators: Energy intensity and energy per capita. The first one, calculated as units of energy per unit of gross domestic product (GDP), is commonly used to measure the energy efficiency of an economy and it is the main indicator, which represents the decoupling of energy from the GDP. The second one is calculated by dividing the total energy consumption by the population.
Focusing on the residential sector, the trend of final energy consumption (FEC) is analysed in comparison with those of the other sectors, and relatively to the member states' groups (both in absolute terms and per capita).
To identify the main influencing determinants on the residential energy consumption, the trends at the EU level and comparisons between member states are presented for several contextual parameters. Considering previous studies [28,[39][40][41][42][43][44] and the data availability, the choice fell on the following ones: Considering that population is an important driver, a discussion of the energy consumption per capita in the EU-28 is first provided. To explain the variation in energy consumption trends between member states, the GDP per capita is also analysed, including the percentage of people without the ability to keep their home adequately warm in the population below the 60% of median equivalised disposable income in the EU-28 Member States. In addition, the adjusted gross disposable income analysis is considered, as it can be a more representative economic indicator for the analysis of energy consumption in residential sector. According to Eurostat definitions [45], the adjusted gross disposable income of a household includes all income from work and from investments and property, transfers between households as well as social transfers. To consider the influence of variations in household size and composition, the total disposable household income is divided by an equivalisation factor. The adjustment improves the comparison of income levels between member states by considering different levels of government involvement in the provision of free services to households.
The importance of the household conditions as an influencing factor is also studied by analysing the household size expressed as the average persons per household in the EU-28 member states. A causal relationship is expected between the number of persons per household and the use of equipment and appliances. In addition, the average floor area might be another significant indicator of households' energy consumption. The larger the size of dwellings, the higher the heating and cooling needs as well as the use of lighting equipment.
Heating degree days (HDD) and cooling degree days (CDD), indicators which are related to the heating and cooling needs, are also studied. A degree-day is defined by Eurostat [45] as a weather-based technical index designed to describe the need for the heating or cooling energy requirements of buildings. They are often used [46] to evaluate the influence of climate and weather conditions in energy consumption trends. In this case, it is studied by analysing the residential energy consumption per dwelling in comparison with the heating degree days (HDD).
The importance of the impact of some of these determinants on energy indicators is studied using linear regression correlation diagrams. This approach can only allow general speculations (i.e., predictions are not generated) because high R-squared values are not expected for this field of application. In fact, the energy consumption of a sector as wide and articulated as the residential one depends heavily on the specific peculiarities of the national building stock and on social and cultural aspects that cannot be taken into account by an article like this [21,27]. Firstly, the correlation between the energy consumption per capita and per average floor area and the heating degree days is analysed. The correlation between the GDP per capita and energy consumption per capita as well as the correlation between the gross adjusted disposable income and energy consumption per capita are also studied.
To understand the trends in residential energy consumption in relation to some important determinants, such as climatic conditions, dwelling-related characteristics, and economic conditions, the authors propose the new indicator, sFEC c,e . This is expressed as the specific residential FEC (sFEC)-that is, the final energy consumption per unit of floor area-corrected for both climatic and economic conditions. The subscripts, "c" and "e", denote the climatic and economic corrections, respectively. The climatic correction is represented by the ratio of the actual heating degree days (HDD) divided by the mean heating degree days over the reference period of 2000-2016 (HDD re f ), while the economic correction is given by the adjusted gross disposable income in the purchasing power standard (GDI) divided by the total number of dwellings (DW). The time period of 2000-2015 was selected because of the availability of data. The results for this indicator are produced using the following equation: The indicator was calculated using the Eurostat datasets of energy balances (Eurostat code: nrg_110a), adjusted gross disposable income per capita (Eurostat code: nasa_10_nf_tr), heating degree days (Eurostat code: nrg_chdd_a), and the Odyssee data on the average floor area per dwelling. The indicator enables the assessment of residential energy consumption trends after the aforementioned important corrections are made, thereby drawing a closer and more precise link between the evolution of the energy consumption trends and the impact of technological and policy effects. Finally, to quantify the impact of possible various factors, we apply decomposition analysis on recent EU-28 residential energy consumption changes in the period of 2005-2016. Decomposition analysis has been widely used by various national and international organisations and agencies [47][48][49] as a tool to inform policy makers on the driving forces behind energy trends [50][51][52][53][54]. Through the use of decomposition analysis, the impact of pre-defined factors-e.g., economic activity, structural shifts in the economy, and weather fluctuations-on changes in energy consumption is determined, thereby enabling the isolation of real energy efficiency trends from other influencing forces. Two of the most popular types of decomposition techniques include the index decomposition analysis (IDA) and structural decomposition analysis (SDA). The main difference between these two types lies in the input data used: The SDA method uses the input-output model to decompose the evolution of indicators, whereas the IDA uses only sectoral data. Among the different IDA methods, the Logarithmic Mean Divisia Index (LMDI-I) carries multiple advantages and was therefore selected as the preferred decomposition technique for this analysis [55,56]. These, inter-alia, include perfect decomposition (that is, the results do not contain any residual term), the possibility to investigate the effect of more than two factors, and the simple relationship between multiplicative and additive forms. The additive form decomposes the difference between two points in time, while the multiplicative form decomposes the ratio of change with respect to the base year. For the residential sector, factors associated with population, wealth, intensity, and weather effects are analysed for each EU member using the hybrid model proposed by Xu and Ang [57]. Due to data restrictions, two subsectors are considered by the authors: (1) Space heating, and (2) all other end-uses. In addition to the population effect (POP), the wealth effect is also studied. The wealth effect is represented by the total floor area of dwellings (TFA) per capita for the end use of space heating and adjusted gross disposable income in purchasing power standard per capita (GDI) for all other end uses. The climatic effect is defined as the ratio of the heating degree days of a specific year (HDD) over the mean heating degree days (HDD re f ) in the reference period of 1990-2016. The climatic adjustment is considered only for the final energy consumption attributed to the space heating end use (FEC h ), while the share of the consumption associated with all other uses (FEC o ) remains unchanged. The decomposition is carried out using the following factorisation identity: where FEC h c stands for the climate corrected final energy consumption for space heating, calculated by dividing the final energy consumption for space heating with the climatic factor: The member state results are added up to give the corresponding effect at the EU level. All applications are run using Eurostat data, except the data on floor area, which are derived from the Odyssee database, as before. The decomposition formulae presented by Ang [58] are used to conduct the analysis.

The Macro Picture: EU Primary and Final Energy Consumption
As stipulated in the EED, primary energy consumption (PEC) covers the consumption of the energy sector, distribution and transformation losses, and the final end-user consumption. It does not include the energy used for non-energy purposes. In 2016, the main component of primary energy consumption is the final energy consumption that accounts for 71.5% of primary energy; followed by transformation losses (21.4%) and consumption in the energy sector (5.2%).  By analysing the most significant energy indicators, such as energy intensity (the ratio between the final energy consumption and gross domestic product (GDP) calculated as chain linked volumes with 2010 as a reference year) and energy per capita, it is possible to observe that from 2000 to 2016, the EU-28 energy intensity dropped by 0.02 toe/thousand Euro, reaching a value of 0.08 toe/thousand Euro in 2016. This reduction is the result of various factors, such as structural economic changes and technological improvements, together with the positive impact of energy efficiency policies, both at the EU and national level. After reaching the lowest value in 2014 (2.10 toe/capita), energy per capita was 2.17 toe/capita in 2016. Given the mild evolution of population over the examined period, energy per capita follows closely the trend of final energy consumption.  By analysing the most significant energy indicators, such as energy intensity (the ratio between the final energy consumption and gross domestic product (GDP) calculated as chain linked volumes with 2010 as a reference year) and energy per capita, it is possible to observe that from 2000 to 2016, the EU-28 energy intensity dropped by 0.02 toe/thousand Euro, reaching a value of 0.08 toe/thousand Euro in 2016. This reduction is the result of various factors, such as structural economic changes and technological improvements, together with the positive impact of energy efficiency policies, both at the EU and national level. After reaching the lowest value in 2014 (2.10 toe/capita), energy per capita was 2.17 toe/capita in 2016. Given the mild evolution of population over the examined period, energy per capita follows closely the trend of final energy consumption. By analysing the most significant energy indicators, such as energy intensity (the ratio between the final energy consumption and gross domestic product (GDP) calculated as chain linked volumes with 2010 as a reference year) and energy per capita, it is possible to observe that from 2000 to 2016, the EU-28 energy intensity dropped by 0.02 toe/thousand Euro, reaching a value of 0.08 toe/thousand Euro in 2016. This reduction is the result of various factors, such as structural economic changes and technological improvements, together with the positive impact of energy efficiency policies, both at the EU and national level. After reaching the lowest value in 2014 (2.10 toe/capita), energy per capita was 2.17 toe/capita in 2016. Given the mild evolution of population over the examined period, energy per capita follows closely the trend of final energy consumption.   In 2016, the residential sector consumed 25.7% of the final energy consumption in the EU. This makes it the second largest consuming sector after transport (33.2%). The increase of the final energy consumption in the transport and services sectors, combined with the decrease in the industrial sector, might be the result of the current tertiarization process in the EU. In 2016, however, the final energy consumption had increased in all sectors compared to the year, 2015, but mostly in the residential sector (3.1%).

Energy Consumption and Energy Efficiency Trends in the Residential Sector
Among the various sectors, the strongest fluctuations are noted in the residential sector, with significant consumption drops of 11.2% and 12.0 % in 2011 and 2014, respectively. Given that two of the warmest winters over this period were recorded in 2011 and 2014, a trend confirmed by the analysis of the heating degree days in Figure 13, it can be concluded that the climatic conditions have a strong impact on the residential energy consumption. Conversely, 2010 represented one of the coldest years, largely explaining the increase in consumption registered in that year. The consumption peak in 2010 (320.0 Mtoe) and the dip in 2014 (265.1 Mtoe), in fact, represent the maximum and minimum consumption levels experienced in a period of a longer span that stretches over 27 years, from 1990 to 2016. As shown in Figure 3, the residential FEC in the EU-28 decreased from 290.9 Mtoe in 2000 to 284.8 Mtoe in 2016. This is equivalent to a drop of 2.1%. consumption in the transport and services sectors, combined with the decrease in the industrial sector, might be the result of the current tertiarization process in the EU. In 2016, however, the final energy consumption had increased in all sectors compared to the year, 2015, but mostly in the residential sector (3.1%).
Among the various sectors, the strongest fluctuations are noted in the residential sector, with significant consumption drops of 11.2% and 12.0 % in 2011 and 2014, respectively. Given that two of the warmest winters over this period were recorded in 2011 and 2014, a trend confirmed by the analysis of the heating degree days in Figure 13, it can be concluded that the climatic conditions have a strong impact on the residential energy consumption. Conversely, 2010 represented one of the coldest years, largely explaining the increase in consumption registered in that year. The consumption peak in 2010 (320.0 Mtoe) and the dip in 2014 (265.1 Mtoe), in fact, represent the maximum and minimum consumption levels experienced in a period of a longer span that stretches over 27 years, from 1990 to 2016. As shown in Figure 3, the residential FEC in the EU-28 decreased from 290.9 Mtoe in 2000 to 284.8 Mtoe in 2016. This is equivalent to a drop of 2.1%.

Potential Determinants Influencing Residential Energy Consumption
The analysis of the influencing factors can provide new insights into the study of energy consumption trends. In the case of the residential sector, it is possible to link energy consumption to specific variables, such as population, GDP per capita, adjusted disposable income, heating degree days, cooling degree days, number of dwellings, and average floor area. The mentioned variables were combined in different ways in an attempt to understand whether the behaviour of the residential FEC is still the same after correcting for these factors. Forecasts that go beyond the available data are not proposed, but general indications are provided, which are based on the assessment of energy consumption trends and the aforementioned determinants.

Population
In 2016, the EU-28 population increased by 4.7% compared to 2000 while in the same period, the residential FEC declined by 2.1%, registering a positive CAGR by 0.3% and a negative CAGR by 0.1%, respectively.

Potential Determinants Influencing Residential Energy Consumption
The analysis of the influencing factors can provide new insights into the study of energy consumption trends. In the case of the residential sector, it is possible to link energy consumption to specific variables, such as population, GDP per capita, adjusted disposable income, heating degree days, cooling degree days, number of dwellings, and average floor area. The mentioned variables were combined in different ways in an attempt to understand whether the behaviour of the residential FEC is still the same after correcting for these factors. Forecasts that go beyond the available data are not proposed, but general indications are provided, which are based on the assessment of energy consumption trends and the aforementioned determinants.

Population
In 2016, the EU-28 population increased by 4.7% compared to 2000 while in the same period, the residential FEC declined by 2.1%, registering a positive CAGR by 0.3% and a negative CAGR by 0.1%, respectively.  The residential FEC per capita declined by 38.9 koe in the EU-28 in the period of 2000-2016, representing a drop of 6.5% (a drop of 0.5% at CAGR). It can be observed that trends in residential FEC have a stronger impact on the trends of this indicator compared to the trends in population, as there are no significant changes in the population during the studied period. Consequently, despite the growth of the EU-28 population, its impact on the residential FEC is limited. FEC have a stronger impact on the trends of this indicator compared to the trends in population, as there are no significant changes in the population during the studied period. Consequently, despite the growth of the EU-28 population, its impact on the residential FEC is limited. The residential FEC per capita declined by 38.9 koe in the EU-28 in the period of 2000-2016, representing a drop of 6.5% (a drop of 0.5% at CAGR). It can be observed that trends in residential FEC have a stronger impact on the trends of this indicator compared to the trends in population, as there are no significant changes in the population during the studied period. Consequently, despite the growth of the EU-28 population, its impact on the residential FEC is limited.

Economic Conditions
Other factors that may have an impact on energy consumption are the economic growth and the economic conditions of the countries. GDP per capita in the EU-28 ( Figure 7) has been continuously growing in the period from 2000 to 2016 except for a single fall in 2009 due to the economic and financial crisis. While GDP per capita increased by 47.5% (registering an increase by 2.4% at CAGR) over the period of 2000-2016, residential FEC per capita fell by 6.5% in the same period. It is observed that there is a large dispersion of GDP per capita in the EU-28, so it is important to look at the disaggregated data per country. Bulgaria registered the lowest GDP per capita (EUR 6728) and Luxembourg had the highest (EUR 91,982) in the face of an average GDP per capita in the EU-28 equal to EUR 29,241. It is easy to notice that Finland, Luxembourg, and Denmark, is a large dispersion of GDP per capita in the EU-28, so it is important to look at the disaggregated data per country. Bulgaria registered the lowest GDP per capita (EUR 6728) and Luxembourg had the highest (EUR 91,982) in the face of an average GDP per capita in the EU-28 equal to EUR 29,241. It is easy to notice that Finland, Luxembourg, and Denmark, which are the countries with the highest consumption per capita, also have above average GDP per capita values. These data could suggest that higher GDP levels can result in the possibility of buying more energy, by having equipment at home that consumes more energy. In this point, it is important to note that residential energy consumption, especially in economically developed countries, is also influenced by the behavioural patterns and cultural habits of the building occupants. On the other hand, Bulgaria and Romania, which are the member states with the minimum GDP per capita, are also member states with low residential FEC per capita values. This could suggest that lower GDP levels may lead to the inability of many households to ensure the required levels of energy in the home, a condition commonly called energy poverty [59].
Indeed, Figure 8 shows that many member states with GDP per capita levels below the EU-28 average and low final residential energy consumption values register high percentages of people without the ability to maintain a warm home in the total low income population. For example, the share of the population without the ability to keep their home adequately warm in the population below the 60% of median equivalised disposable income was above 50% in Bulgaria and Greece (61.9% and 52.5%, respectively).
On the other hand, Luxemburg and Denmark, both member states with high values of final residential energy consumption values, but also high GDP per capita values, registered low percentages in the low income population (4.0% and 7.9%, respectively) of people not able to ensure the energy required for heating needs. Overall, there is a slight decrease in the share of people not able to keep their home warm in the low income population (by 0.1%) in EU-28 from 2010 to 2016. Possibly, there is probably an inverse correlation between this indicator and GDP per capita.    The growth in the GDP per capita, despite the population increase, can be the result of significant economic developments during recent years. The comparison of economic and energy consumption trends indicates that important economic growth has not been accompanied by a growth of the same magnitude in energy consumption. The adjusted gross disposable income per capita can be considered more relevant to the residential sector because it describes the wealth of households occupying residential buildings. In fact, by comparing the trend-lines of both adjusted gross disposable income per capita and GDP per capita for the 17-year period from 2000 to 2016 (Figure 9), a notable difference between these two variables can be observed. In 2016, the difference reached EUR 7347. Thus, the GDP, although largely used, if adopted in the analysis of the residential The growth in the GDP per capita, despite the population increase, can be the result of significant economic developments during recent years. The comparison of economic and energy consumption trends indicates that important economic growth has not been accompanied by a growth of the same magnitude in energy consumption. The adjusted gross disposable income per capita can be considered more relevant to the residential sector because it describes the wealth of households occupying residential buildings. In fact, by comparing the trend-lines of both adjusted gross disposable income per capita and GDP per capita for the 17-year period from 2000 to 2016 (Figure 9), a notable difference between these two variables can be observed. In 2016, the difference reached EUR 7347. Thus, the GDP, although largely used, if adopted in the analysis of the residential sector, may not be a representative indicator for end-users as it does not reflect the real purchasing power of the inhabitants.   Figure 10 shows that the member states with the highest adjusted gross disposable income levels (Luxemburg, Austria, Belgium) are also among the countries with the highest final residential energy consumption per capita for the year of 2016. On the contrary, Bulgaria, which is the member state with the lowest adjusted gross disposable income per capita, also has the third lowest residential FEC per capita value after Malta and Portugal. This indicates that there is a correlation between the final residential energy consumption per capita and adjusted gross disposable income. However, dwelling-related and climatic corrections need to be applied to the final energy consumption per capita to assess whether there is a correlation between economic development and energy consumption trends. From this graph, Malta and Croatia were removed, as there are no available data for their adjusted gross disposable income for 2016.  Figure 10 shows that the member states with the highest adjusted gross disposable income levels (Luxemburg, Austria, Belgium) are also among the countries with the highest final residential energy consumption per capita for the year of 2016. On the contrary, Bulgaria, which is the member state with the lowest adjusted gross disposable income per capita, also has the third lowest residential FEC per capita value after Malta and Portugal. This indicates that there is a correlation between the final residential energy consumption per capita and adjusted gross disposable income. However, dwelling-related and climatic corrections need to be applied to the final energy consumption per capita to assess whether there is a correlation between economic development and energy consumption trends. From this graph, Malta and Croatia were removed, as there are no available data for their adjusted gross disposable income for 2016. Figure 10 shows that the member states with the highest adjusted gross disposable income levels (Luxemburg, Austria, Belgium) are also among the countries with the highest final residential energy consumption per capita for the year of 2016. On the contrary, Bulgaria, which is the member state with the lowest adjusted gross disposable income per capita, also has the third lowest residential FEC per capita value after Malta and Portugal. This indicates that there is a correlation between the final residential energy consumption per capita and adjusted gross disposable income. However, dwelling-related and climatic corrections need to be applied to the final energy consumption per capita to assess whether there is a correlation between economic development and energy consumption trends. From this graph, Malta and Croatia were removed, as there are no available data for their adjusted gross disposable income for 2016.  Figure 11. The overall trend in Europe is the result of population growth in combination with the rise   Figure 12 shows that between 2000 and 2015, the average floor area of EU-28 dwellings did not change considerably (increased by 5.1 m 2 or almost 6% in the whole period). The drop of 22% in the consumption per dwelling climatic corrected during the same period, which was calculated by Odyssee, thus cannot be explained by a higher population density or smaller houses. It might be the result of a higher share of more efficient equipment and appliances, and other improvements in building elements (e.g., envelopes and thermal systems). However, weather conditions may also  Odyssee, thus cannot be explained by a higher population density or smaller houses. It might be the result of a higher share of more efficient equipment and appliances, and other improvements in building elements (e.g., envelopes and thermal systems). However, weather conditions may also have played an important role. The negative correlation between the two variables shown in this graph (estimated at around −0.98) is probably an indication of the fact that the observed progressive increase of floor areas has been accompanied by a decrease in energy consumption per unit of floor area of dwellings. This type of correlation has already been observed in the existing literature, mostly in relation to energy consumption for heating [60,61], and might be explained in terms of a higher energy efficiency resulting from smaller 'surface-area-to-volume' ratios (and hence heat losses) of larger dwellings. Further and more detailed studies would, however, be needed to verify this hypothesis on the data set used in the present analysis. The time period studied in this graph does not include 2016 as there are no available data for this year.  Figure 12 shows that between 2000 and 2015, the average floor area of EU-28 dwellings did not change considerably (increased by 5.1 m 2 or almost 6% in the whole period). The drop of 22% in the consumption per dwelling climatic corrected during the same period, which was calculated by Odyssee, thus cannot be explained by a higher population density or smaller houses. It might be the result of a higher share of more efficient equipment and appliances, and other improvements in building elements (e.g., envelopes and thermal systems). However, weather conditions may also have played an important role. The negative correlation between the two variables shown in this graph (estimated at around −0.98) is probably an indication of the fact that the observed progressive increase of floor areas has been accompanied by a decrease in energy consumption per unit of floor area of dwellings. This type of correlation has already been observed in the existing literature, mostly in relation to energy consumption for heating [60,61], and might be explained in terms of a higher energy efficiency resulting from smaller 'surface-area-to-volume' ratios (and hence heat losses) of larger dwellings. Further and more detailed studies would, however, be needed to verify this hypothesis on the data set used in the present analysis. The time period studied in this graph does not include 2016 as there are no available data for this year.

Climatic Conditions
Environmental conditions, such as weather and climate, can affect energy consumption: For example, a cold winter or a hot summer can result in occasional consumption peaks. The residential FEC per dwelling has been dropping between 2000 and 2016 in the EU-28. In 2000, the residential consumption per dwelling was 1.56 toe. In 2016, the consumption per dwelling was 1.32 toe, corresponding to a fall of 18.2% (1% at CAGR). It is important to note that the final residential energy consumption per dwelling was calculated by the authors by dividing the final residential energy consumption by the stock of permanently occupied dwellings in the EU-28 as made available by Odyssee. Part of this decrease can be explained assuming that there is a correlation between the computed energy consumption per dwelling and the changing climatic conditions. A comparison between the total final energy consumption per dwelling and heating degree days is shown in Figure 13.

Correlations
The following set of four figures aims to investigate how residential consumption is correlated with various climatic and economic conditions. The analysis provides some evidence about the strength and direction of the linear relationship between our variables of interest, and it does not have any predictive claim. Figure 14 shows the correlation between residential energy consumption for space heating per capita and per m 2 and HDD in the EU-28 member states for the year of 2015. The selection of the year studied in this graph was made based on the availability of the data. Belgium and Romania were removed from the final results due to incomplete data. The first evidence is that, as expected, countries with a colder climate have, on average, higher levels of residential consumption. The

Correlations
The following set of four figures aims to investigate how residential consumption is correlated with various climatic and economic conditions. The analysis provides some evidence about the strength and direction of the linear relationship between our variables of interest, and it does not have any predictive claim. Figure 14 shows the correlation between residential energy consumption for space heating per capita and per m 2 and HDD in the EU-28 member states for the year of 2015. The selection of the year studied in this graph was made based on the availability of the data. Belgium and Romania were removed from the final results due to incomplete data. The first evidence is that, as expected, countries with a colder climate have, on average, higher levels of residential consumption. The second and more interesting evidence is that countries with the same climatic conditions can perform very differently in terms of residential consumption. Eastern European countries show a higher level of residential consumption with respect to the average. Among the countries with a colder climate, Finland and Sweden provide examples of a relatively low level of residential energy consumption. The behavioural, cultural and social habits, and lifestyle also have an impact on the energy consumption patterns in these countries. Nevertheless, it is impossible to perform a quantitative assessment of this impact due to the lack of data and, if available, due to the nature of these data, which are often not comparable among countries. The value of the R squared in Figure 14 is around 0.55. This confirms that a linear relationship between energy consumption per capita and per m 2 and HDD exists and is positive, in line with our hypothesis. Figure 15 shows the correlation between residential final energy consumption per capita and GDP per capita in the EU-28 member states for the year of 2016. From the figure, it would seem that in some cases, countries with higher GDP per capita values (i.e., countries with better economic conditions) tend to consume more energy in the residential sector. However, there are notable differences in energy consumption between member states with similar levels of GDP per capita. Regarding countries below the trend-line with relatively high GDP per capita values, but also with increased heating needs due to their weather conditions, such as Luxemburg (2967.47 HDD, above EU-28 HDD average), it is possible to argue that they might have made more efficient use of energy sources. Similar conclusions can be gathered from Figure 16, in which a slightly better correlation between the final residential energy per capita and adjusted disposable income (with a lower distance between the fitted line and all of the data points) is presented. This confirms that adjusted gross disposable income is a more representative economic indicator compared to GDP for the analysis of the residential energy consumption trends. It should be noted that Croatia and Malta were eliminated from Figure 16 due to incomplete data. Looking at both figures, it is possible to observe that some points are close to the line, but other points are far from it, which indicates only a moderate linear relationship between the variables. It is also worth noticing that correlation coefficients are very sensitive to extreme data values. Even if the R squared value is below 0.5 in both cases, the slope of the regression lines is evidence showing the tendency of economically developed countries to consume more energy. Moving beyond a simple correlation by adding other explanatory variables would allow us to better capture the variability among data and to explain causality. However, inferring causality is out of the scope of this article and we lack some relevant behavioural variables that are crucial to explaining residential consumption patterns. second and more interesting evidence is that countries with the same climatic conditions can perform very differently in terms of residential consumption. Eastern European countries show a higher level of residential consumption with respect to the average. Among the countries with a colder climate, Finland and Sweden provide examples of a relatively low level of residential energy consumption. The behavioural, cultural and social habits, and lifestyle also have an impact on the energy consumption patterns in these countries. Nevertheless, it is impossible to perform a quantitative assessment of this impact due to the lack of data and, if available, due to the nature of these data, which are often not comparable among countries. The value of the R squared in Figure  14 is around 0.55. This confirms that a linear relationship between energy consumption per capita and per m 2 and HDD exists and is positive, in line with our hypothesis.  Figure 15 shows the correlation between residential final energy consumption per capita and GDP per capita in the EU-28 member states for the year of 2016. From the figure, it would seem that in some cases, countries with higher GDP per capita values (i.e., countries with better economic conditions) tend to consume more energy in the residential sector. However, there are notable differences in energy consumption between member states with similar levels of GDP per capita. Regarding countries below the trend-line with relatively high GDP per capita values, but also with increased heating needs due to their weather conditions, such as Luxemburg (2967.47 HDD, above EU-28 HDD average), it is possible to argue that they might have made more efficient use of energy sources. Similar conclusions can be gathered from Figure 16, in which a slightly better correlation between the final residential energy per capita and adjusted disposable income (with a lower distance between the fitted line and all of the data points) is presented. This confirms that adjusted gross disposable income is a more representative economic indicator compared to GDP for the analysis of the residential energy consumption trends. It should be noted that Croatia and Malta were eliminated from Figure 16 due to incomplete data. Looking at both figures, it is possible to observe that some points are close to the line, but other points are far from it, which indicates only a moderate linear relationship between the variables. It is also worth noticing that correlation coefficients are very sensitive to extreme data values. Even if the R squared value is below 0.5 in both cases, the slope of the regression lines is evidence showing the tendency of economically developed countries to consume more energy. Moving beyond a simple correlation by adding other explanatory variables would allow us to better capture the variability among data and to explain causality. However, inferring causality is out of the scope of this article and we lack some relevant behavioural variables that are crucial to explaining residential consumption patterns.   The analysis conducted so far demonstrates that every determinant discussed above influences in a certain way (but never in a definitive way) the residential final energy consumptions. Hence, it is plausible to try to compare the national values considering all the aforementioned influential factors in order to derive general considerations about how the energy policies implemented resulted in a more efficient use of energy. This could be achieved by calculating the final energy consumption normalized by indicators corresponding to the mentioned factors (HDD, average floor area, and population). Figure 17 shows that the degree of correlation between the adjusted gross disposable income per capita and the climatic corrected residential final energy consumption per capita and per average floor area is very low. A low correlation coefficient does not necessarily imply that no The analysis conducted so far demonstrates that every determinant discussed above influences in a certain way (but never in a definitive way) the residential final energy consumptions. Hence, it is plausible to try to compare the national values considering all the aforementioned influential factors in order to derive general considerations about how the energy policies implemented resulted in a more efficient use of energy. This could be achieved by calculating the final energy consumption normalized by indicators corresponding to the mentioned factors (HDD, average floor area, and population). Figure 17 shows that the degree of correlation between the adjusted gross disposable income per capita and the climatic corrected residential final energy consumption per capita and per average floor area is very low. A low correlation coefficient does not necessarily imply that no relationship exists, but it simply suggests that this relationship is not linear. Consequently, based on this evidence, it is difficult to derive conclusions regarding the impact of disposable income on the "corrected" residential energy consumption trends. However, as before, we can interpret the negative correlation. The change in the sign of the correlation is important as it shows that the results change significantly when climatic data are applied. Indeed, countries with a colder climate, such as Denmark, Sweden, and Finland, are found below the trendline in this graph. To apply the climatic correction, the final residential energy consumption was divided by the climatic factor; that is, the heating degree days of a member state for the year of 2015 divided by the mean heating degree days of the EU-28 member states for the same year. 2015 was selected based on the availability of data. It should be noted that Malta, Croatia, and Belgium were removed from the graph due to incomplete data. such as Denmark, Sweden, and Finland, are found below the trendline in this graph. To apply the climatic correction, the final residential energy consumption was divided by the climatic factor; that is, the heating degree days of a member state for the year of 2015 divided by the mean heating degree days of the EU-28 member states for the same year. 2015 was selected based on the availability of data. It should be noted that Malta, Croatia, and Belgium were removed from the graph due to incomplete data. As discussed in Section 3, this represents the final residential energy consumption adjusted to take into account all the factors mentioned above (adjusted gross disposable income per dwelling, climatic conditions, and floor area of dwellings). The climatic correction was applied by dividing the heating degree days for a specific year by the mean heating degree days for the period of 2000-2016.
The results show a steady decline over the examined period, which resulted in an , drop of nearly 40% (or 3.4% at CAGR) in 2015 compared to 2000 (from 0.48 ktoe/m 2 /thousand EUR in 2000 to 0.29 ktoe/m 2 /thousand EUR in 2015). As a result of the corrections made, a more linear , trend is observed in Figure 18. While the evolution of residential without any corrections does not follow any specific path (a very low R 2 value of 0.26 is given for the linear relationship between residential and time in Figure 3), the evolution of , over time shown in Figure 18 can be described by a clear linear relationship with an R 2 value of 0.96. By making all the aforementioned adjustments, we are thus capable of removing the yearly residential fluctuations that are associated with variations in economic, climatic, and dwelling conditions. Our analysis indicates that at the EU level, the impact of technological and policy changes over the period of 2000-2015 resulted in a constant annual , decline rate of 0.0127 ktoe/m 2 /thousand EUR ( Figure 18). As discussed in Section 3, this represents the final residential energy consumption adjusted to take into account all the factors mentioned above (adjusted gross disposable income per dwelling, climatic conditions, and floor area of dwellings). The climatic correction was applied by dividing the heating degree days for a specific year by the mean heating degree days for the period of 2000-2016. The results show a steady decline over the examined period, which resulted in an sFEC c,e drop of nearly 40% (or 3.4% at CAGR) in 2015 compared to 2000 (from 0.48 ktoe/m 2 /thousand EUR in 2000 to 0.29 ktoe/m 2 /thousand EUR in 2015). As a result of the corrections made, a more linear sFEC c,e trend is observed in Figure 18. While the evolution of residential FEC without any corrections does not follow any specific path (a very low R 2 value of 0.26 is given for the linear relationship between residential FEC and time in Figure 3), the evolution of sFEC c,e over time shown in Figure 18 can be described by a clear linear relationship with an R 2 value of 0.96. By making all the aforementioned adjustments, we are thus capable of removing the yearly residential FEC fluctuations that are associated with variations in economic, climatic, and dwelling conditions. Our analysis indicates that at the EU level, the impact of technological and policy changes over the period of 2000-2015 resulted in a constant annual sFEC c,e decline rate of 0.0127 ktoe/m 2 /thousand EUR ( Figure 18). Figure 19 shows the specific residential FEC corrected for climatic and economic conditions for each member state. To harmonise geographical-related climatic variations in our analysis, the climatic correction was expressed as the ratio of the actual heating degree days of a given country and the mean heating degree days over the reference period of the EU. As shown in Figure 19, a considerably different energy efficiency ranking of the member states is given compared to Figure 5. By introducing climatic, economic, and dwelling-related corrections, Finland, Sweden, Austria, and Luxembourg-some of the highest consumers of energy based on our Figure 5 analysis-are now the four countries with the lowest level of final residential energy consumption. At the same time, countries that performed very well in terms of absolute values of final energy consumption are ranked among the countries that consume the most according to Figure 19. It should be noted that Malta was excluded from the figure due to the unavailability of data.  Figure 19 shows the specific residential FEC corrected for climatic and economic conditions for each member state. To harmonise geographical-related climatic variations in our analysis, the climatic correction was expressed as the ratio of the actual heating degree days of a given country and the mean heating degree days over the reference period of the EU. As shown in Figure 19, a considerably different energy efficiency ranking of the member states is given compared to Figure 5. By introducing climatic, economic, and dwelling-related corrections, Finland, Sweden, Austria, and Luxembourg-some of the highest consumers of energy based on our Figure 5 analysis-are now the four countries with the lowest level of final residential energy consumption. At the same time, countries that performed very well in terms of absolute values of final energy consumption are ranked among the countries that consume the most according to Figure 19. It should be noted that Malta was excluded from the figure due to the unavailability of data.   Figure 19 shows the specific residential FEC corrected for climatic and economic conditions for each member state. To harmonise geographical-related climatic variations in our analysis, the climatic correction was expressed as the ratio of the actual heating degree days of a given country and the mean heating degree days over the reference period of the EU. As shown in Figure 19, a considerably different energy efficiency ranking of the member states is given compared to Figure 5. By introducing climatic, economic, and dwelling-related corrections, Finland, Sweden, Austria, and Luxembourg-some of the highest consumers of energy based on our Figure 5 analysis-are now the four countries with the lowest level of final residential energy consumption. At the same time, countries that performed very well in terms of absolute values of final energy consumption are ranked among the countries that consume the most according to Figure 19. It should be noted that Malta was excluded from the figure due to the unavailability of data.

Decomposition Analysis
An alternative way to quantify the influence of the mentioned factors is the decomposition analysis that was applied in the EU-28 for the period of 2005-2016. The multiplicative decomposition results of the final energy consumption of the EU-28 in the residential sector are shown in Figure 20. The decline of 8% (equivalent to 24.6 Mtoe) in EU-28 residential consumption over the examined period was primarily driven by improvements in energy intensity-which contributed to a reduction of 61.4 Mtoe in this sector, equivalent to a drop of 20% compared to 2005 consumption levels-and, to a lesser extent, warmer winters, which were associated with a drop of energy consumption by 13.1 Mtoe in 2016 compared to 2005 levels (4.2%). As a counteracting effect, activity, measured both in terms of changes in population and wealth, drove up consumption at the EU level: The population effect was responsible for a 9.9 Mtoe hike in consumption (3.2%) and the wealth effect caused an increase of 40 Mtoe (13.0%). The activity effect has been constantly rising in the period of 2005-2016, while the opposite is true for the intensity effect. It can also be seen that the weather effect trend is closely correlated with the total consumption trend, confirming the strong impact of weather fluctuations on the total residential energy consumption from our indicator-based analysis. This provides the main explanation behind the dip in consumption in 2011 and 2014, and, conversely, the hike in consumption in 2010. The recorded EU heating degree days in the years of 2011 and 2014 were 2953 and 2809, 10% and 5% higher than the average 3181 heating degree days over the period of 1990-2016, while 2010 was a cold year with a recorded 3485 degree days (11% above the average). The decomposition results show that by removing the population, wealth, and weather effects from the consumption trends, a clearer picture of the impact of policy and technological changes can be obtained. It can also be observed that the trend of the intensity effect seems to be closely interrelated with the sFEC c,e trend in the overlapped period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). In line with the evolution of sFEC c,e shown in Figure 18, a clear linear path of the intensity effect is also obtained in Figure 20. This confirms the positive impact of technological and policy changes during the examined period. Our results, for the period of 2005-2016, obtained by using the multiplicative method, are in good agreement with previously published results [34]. However, it is important to note that our results are not directly comparable with those of the study mentioned above, as Reuter et al. analysed the period of 2000-2015 by using the additive method.
decomposition results of the final energy consumption of the EU-28 in the residential sector are shown in Figure 20. The decline of 8% (equivalent to 24.6 Mtoe) in EU-28 residential consumption over the examined period was primarily driven by improvements in energy intensity-which contributed to a reduction of 61.4 Mtoe in this sector, equivalent to a drop of 20% compared to 2005 consumption levels-and, to a lesser extent, warmer winters, which were associated with a drop of energy consumption by 13.1 Mtoe in 2016 compared to 2005 levels (4.2%). As a counteracting effect, activity, measured both in terms of changes in population and wealth, drove up consumption at the EU level: The population effect was responsible for a 9.9 Mtoe hike in consumption (3.2%) and the wealth effect caused an increase of 40 Mtoe (13.0%). The activity effect has been constantly rising in the period of 2005-2016, while the opposite is true for the intensity effect. It can also be seen that the weather effect trend is closely correlated with the total consumption trend, confirming the strong impact of weather fluctuations on the total residential energy consumption from our indicator-based analysis. This provides the main explanation behind the dip in consumption in 2011 and 2014, and, conversely, the hike in consumption in 2010. The recorded EU heating degree days in the years of 2011 and 2014 were 2953 and 2809, 10% and 5% higher than the average 3181 heating degree days over the period of 1990-2016, while 2010 was a cold year with a recorded 3485 degree days (11% above the average). The decomposition results show that by removing the population, wealth, and weather effects from the consumption trends, a clearer picture of the impact of policy and technological changes can be obtained. It can also be observed that the trend of the intensity effect seems to be closely interrelated with the , trend in the overlapped period (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015). In line with the evolution of , shown in Figure 18, a clear linear path of the intensity effect is also obtained in Figure 20. This confirms the positive impact of technological and policy changes during the examined period. Our results, for the period of 2005-2016, obtained by using the multiplicative method, are in good agreement with previously published results [34]. However, it is important to note that our results are not directly comparable with those of the study mentioned above, as Reuter

Conclusions
This article described and analysed the energy consumption patterns in the EU for the period of 2000-2016, with a focus on the residential sector. This analysis aimed to give new insights into the impact and effectiveness of energy efficiency policies implemented in the EU for this sector. The results show that EU primary energy consumption and final energy consumption decreased by 4.6% and 2.2% in 2016 compared to 2000. Energy indicators, such as the final energy intensity and final energy per capita, decreased during the analysed period, by 20.0% and 6.5%, respectively (1.5% and 0.4% at CAGR).
The residential sector registered a final energy consumption drop of 2.1% in the period from 2000 to 2016. However, residential FEC increased by almost 6% from 2000 to 2005, and recorded its maximum in 2010. After the lowest value (over the period of 1990-2016) in 2014, final residential energy consumption increased by 4.3% in 2015 compared to the previous year. Additionally, in 2016, an annual increase of 3.1% was registered. Despite the recent hike of 2015 and 2016, consumption has remained below the average consumption of the last two decades. To better understand the role of the economic crisis in 2008 and the subsequent economic recovery, the adjusted gross disposable income was considered in the analysis, including the evolution of the residential energy consumption compared to the GDP per capita. As shown by the decomposition results, energy consumption in the residential sector dropped despite the upward driving force exerted by the wealth effect linked to economic growth. While the economic recession in 2008-2009 has strongly hit some MSs, its impact was not evident at the EU level as the overall wealth effect (reflected by the higher floor area per inhabitant over time as well as the increase in the adjusted gross disposable income) drove up the residential energy throughout the time period considered in this article.
Weather and climatic conditions, as expected, seem to have a profound impact on the residential energy demand. The results show that despite some exceptions, the colder the year, the higher the energy consumption. Our analysis showed a relatively strong correlation between the final energy consumption and heating degree days. The decomposition results also confirmed this positive correlation by quantifying the impact of climatic variations on the residential energy consumption. According to the decomposition results, the fluctuations in EU residential energy consumption were largely attributed to the climatic variations experienced over the studied period. Nevertheless, establishing a direct impact of climatic conditions on residential energy consumption is not easy, given that several other factors, such as building stock characteristics, social and cultural peculiarities as well as economic conditions, among others, affect the consumption.
Beyond climatic variations, the consideration of sector-specific determinants, such as household and building socio-economic characteristics (e.g., average floor area of dwellings, average number of persons per household), were assessed to gain a more complete picture of the energy efficiency improvements of the residential sector. The determinants used were in line with those adopted by other researchers.
To this end, the new sFEC c,e indicator proposed by the authors offers a better understanding of the residential final energy consumption trends by applying corrections for economic, climatic, and dwelling-related factors. The results showed a steady decline of sFEC c,e over the examined period, corresponding to an overall drop of nearly 40% in 2015 compared to 2000 and an annual decline rate of 0.0127 ktoe/m 2 /1000 EUR. Moreover, our results showed that the decline of sFEC c,e over time can be described by a clear linear relationship. All these findings support a very important conclusion emerging from this article: When residential energy consumption is normalized for the most important drivers described in the article, residential energy consumption is still declining, thus implying that energy efficiency policy must have had a clear role in the decreasing trends. Indeed, this was confirmed also by the decomposition results, according to which declining energy intensity was the main driver behind the overall residential energy consumption over the examined period. To advance the understanding of residential energy demand, econometric modelling can offer additional insights of the actual impact of energy efficiency policies on these trends [21]. Although this study does not allow the impact of energy policy to be isolated completely, it is coherent with the results from other studies [21,22,27].
For what concerns policy implications, the increase in primary and final consumption registered in 2015 and 2016 stresses even more the need for continuing and strengthening policy efforts at the EU and MS level in the remaining years. Policies shall be pursued and reinforced to ensure that the 2020 and 2030 targets are reached. The 2018 EPBD and EED revisions will help, as well as the revision of MEPs adopted under the Ecodesign directive. Whilst the efficiency of appliances is rapidly increasing as a result of the Ecodesign directive, building refurbishments to MEPs are slowly progressing. Against this background, financing is critical and the launch of the Smart Finance for Smart Buildings Initiative [62] can contribute to a more effective use of public funds, to a more complete assistance for the creation of project pipelines, and to a change in the risk perception of financiers and investors.
Further research is needed, including an analysis of the new Eurostat updates covering the 2017 energy data, as well as an assessment of electricity and gas consumption trends. In addition, an analysis of the consumption trends for the various end-uses (energy consumption for space heating, space cooling, lighting and appliances, etc.) as well as a more detailed decomposition analysis may be included in future research. Finally, further econometric analysis may complement the results.
Author Contributions: S.T.T. carried out the main analysis and wrote the first draft of the article. M.E. carried out the decomposition analysis, wrote the related section and performed the English editing. P.B. supervised the whole research project and the article and wrote the Literature Review section. L.C., F.D., N.L., P.Z. and T.R.S. provided critical feedback and helped shape the research, analysis and manuscript.
Funding: This research received no external funding.

Conflicts of Interest:
The authors declare no conflict of interest.