Improving energy poverty measurement in Southern European Regions through equivalisation of theoretical energy costs

sustainability Article Improving Energy Poverty Measurement in Southern European Regions through Equivalization of Modeled Energy Costs Iñigo Antepara 1,* , Lefkothea Papada 2, João Pedro Gouveia 3, Nikolas Katsoulakos 2and Dimitris Kaliampakos 4 1Alokabide, Technical Department, Portal de Gamarra, 1A—2a planta (Edificio el Boulevard), 01013 Vitoria-Gasteiz, Spain 2Metsovion Interdisciplinary Research Center, National Technical University of Athens, 9 Heroon Polytechneiou Str., Zographos, 15780 Athens, Greece; [email protected] (L.P.); [email protected] (N.K.) 3CENSE Center for Environmental and Sustainability Research, NOVA School of Science and Technology, NOVA University Lisbon, 2829-516 Caparica, Portugal; [email protected] 4School of Mining and Metallurgical Engineering, National Technical University of Athens, 9 Heroon Polytechneiou Str., Zographos, 15780 Athens, Greece; [email protected] *Correspondence: [email protected]; Tel.: +34-945-000-565 Received: 21 June 2020; Accepted: 14 July 2020; Published: 16 July 2020   Abstract: In many European countries, energy poverty is measured on the basis of real energy bills, as theoretical energy costs are hard to calculate. The UK is an exception—the data inputs for the Low Income-High Cost (LIHC) indicator are based on reasonable energy costs, these data are collected through specially designed surveys, often an intensive and costly procedure. Approaches which calculate energy needs are valid when energy bill data are unreliable or where households restrict consumption. In this analysis, energy poverty levels are evaluated for Greece, the municipality of É vora (Portugal), and the Basque Country (Spain): energy bills are modeled based on building energy performance data and other energy uses, and adjusted according to socio-demographic variables. To this end, equivalization weights are calculated using socio-economic data from the aforementioned southern European countries/regions. Data are analyzed to compare measurements with actual versus modeled bills using the Ten-Percent Rule (TPR) and Hidden Energy Poverty (HEP) against twice the median (2M) indicator, enhancing the identification of households with low energy consumption. In conclusion, theoretical energy needs can be combined with socio-demographic data instead of actual energy bills to measure energy poverty in a simplified way, avoiding the problem of targeting households that under consume. Keywords: energy poverty; modeled energy costs; TPR; HEP; equivalization coefficients; under consumption; Greece; Portugal; Spain 1. Introduction Between 54 and 141 million people are unable to keep their homes adequately warm in Europe; these citizens are therefore considered to be in energy poverty [ 1 ]. Day et al. [ 2 ] define energy poverty as “insufficient access to affordable, reliable and safe energy services”. The European Commission (EC) is aware of the situation within many European households. In 2016, The EC published the Winter Package, which aimed to shape the clean energy transition, whilst avoiding the marginalization of vulnerable consumers. The second of eight legislative acts approved in 2019 was the European Parliament legislative resolution of 13 November 2018 on the Sustainability 2020,12, 5721; doi:10.3390/su12145721 www.mdpi.com/journal/sustainability Sustainability 2020,12, 5721 2 of 21 proposal for a directive of the European Parliament and of the Council amending Directive 2012/27/EU on energy efficiency, which states that “the Union’s energy efficiency policies should be inclusive and should therefore ensure accessibility to energy efficiency measures for consumers affected by energy poverty”. The EU Energy Poverty Observatory (EPOV) [ 3 ] initiative was launched by the EC to assist Member States (MS) in their efforts to mitigate energy poverty. One of its tasks is to define useful indicators to track the energy poverty situation across Europe. The common indicators proposed by the EPOV are [3]: • Arrears on utility bills, share of (sub-) population having arrears on utility bills, based on self-reported experiences of limited access to energy services (i.e., EU-Statistics on Income and Living Conditions (SILC) data); • Low absolute energy expenditure (M/2), share of households whose absolute energy expenditure is below half the national median, based on household income and/or energy expenditure data (i.e., Housing Budget Survey (HBS) data); • High share of energy expenditure in income (2M), proportion of households whose share of energy expenditure in income is more than twice the national median share, based on HBS data; • Inability to keep home adequately warm, share of (sub)population not able to keep their home adequately warm, based on EU-SILC data. As acknowledged by previous research [ 4 – 6 ], measuring and monitoring energy poverty is challenging. The multidimensionality of the condition means that it is not captured through a single indicator, and there is also a lack of relevant and detailed key data that could be used for cross-country comparisons and for detailed regional analysis. The first attempt to measure energy poverty was applied in the UK, using the Boardman [ 7 ] methodology. This method, known as the Ten-Percent Rule (TPR), calculates the ratio between income and energy costs, where energy costs should not exceed 10% of household income. Energy costs are modeled rather than based on actual bills, and include space heating, domestic hot water (DHW), cooking, lighting, and additional electric appliances [ 8 ]. Critics of the method questioned the means of calculating energy costs [ 6 ] and identified an underrepresentation of the impact of other variables, such as the improvement of energy efficiency [ 9 ]. Additionally, Boardman’s 10% approach was designed for the UK, so its relevance may vary between different countries and income levels [6]. In 2012, the UK revised the definition and adopted the methodology developed by Hills [ 9 ], known as the Low Income-High Cost (LIHC), based on reasonable energy costs. Hills considered two additional criteria to measure energy poverty [ 10 ]: to identify low-income households living in inefficient dwellings as fuel-poor, and a self-reported, subjective index. Hills [ 9 ] finally recommended the energy poverty indicator was based on reasonable energy costs, and in doing so, the LIHC is regarded as a benchmark for ‘objective’ expenditureand income-based indicators [ 11 ]. The main point is that energy poverty centers on a lack of sufficient warmth and space heating [ 12 ], arguably the most important domestic energy service, as a consequence of the well-known implications of living in a cold home for physical and mental health. The basis of the Hills methodology [ 9 ] is that well-insulated dwellings require less energy to maintain a prescribed indoor thermal comfort and this can be accounted for in building energy modeling techniques [10]. Although the Boardman method also uses modeled energy costs instead of actual bills, in the words of Hills, this is the major advantage of this type of methodology [ 10 ]. The main difference between both methods is that Hills proposed to fix the energy poverty threshold at the median energy expenditure. The BRE Domestic Energy Model (BREDEM) modeling tool, using data from the English Housing Survey (EHS), is applied to calculate the threshold of the median modeled bill [ 13 ], avoiding energy costs related to unheated spaces, e.g., due to excessive size of the building, using various heating regimes dependent on the age of the household and employment status, and adjusting for the number of occupants. A standard heating pattern, i.e., assuming that during usual working hours the house Sustainability 2020,12, 5721 3 of 21 is empty, does not apply, in particular for the vulnerable, e.g., the elderly or those caring for young children [ 13 ]. For these cases, the full heating regime was introduced in the UK methodology. Size and number of occupants is also important; therefore, for under-occupied dwellings, partial standard/full heating regimes were included. Initially in the LIHC definition, the addition of non-heating energy uses was not clear [ 14 ]; following the unpacking of the methodology used to generate the UK’s annual fuel poverty statistics these measures became more prominent. The steps involved in this methodology can be summarized as follows: income is measured without considering housing costs, the so-called after housing costs [ 15 ], the dynamic 2M is used for the affordability threshold. Then, the ‘fuel poverty gap’ is added as an additional indicator in order to consider the difference, in monetary terms, between the modeled expenditure for an energy poor household and the expenditure required to remove that household from energy poverty. Subsequently, once modeled energy (and housing) costs are deducted, those whose income is not within the 60 per cent median range are identified as energy poor [10]. Hills [ 9 ] included both energy bills and income details, and identified that they should be adjusted so that comparison to a single threshold is possible for households with different numbers of occupants, known as ‘equivalization’ of income measurement. Equivalization allows comparison on the same scale between households of different composition and size: decreasing the expenses of multiple person households and increasing the expenses of single person households in order to make them comparable. Income equivalization for the Organization for Economic Co-operation and Development (OECD) countries was depicted by [ 16 ]: as 1 head, 0.5 other adult, 0.3 children under 14. In the UK methodology, income is equivalized after the discounting of housing costs [13]. To secure a good level of capabilities, the necessary amount of energy services is dependent on household size and the climate, as previously described, but also specific individuals’ needs and circumstances—e.g., whether they are older, disabled, very young, or ill [ 2 ]. To account for this effect, notwithstanding heating regimes, Hills included the equivalization of energy costs. Initially, using the same factors used for income (adapted from the modified OECD scale), but as the relationship to household of size energy costs and general living costs are not the same, the idea was later rejected [ 9 ]. For instance, if the number of occupants increases by one person in an identical space, heating use would vary very little, while the cost of living would increase [ 9 ]. Furthermore, the energy demands of specific consumers can exceed the average as a result of socio-demographic factors, which can lead to their being identified as vulnerable, i.e., if they experience difficulties when accessing energy [ 17 ]. If those vulnerable characteristics are to be included, additional adjustments are needed, resulting in the equivalized theoretical energy costs. The equivalization factors for energy costs are reviewed periodically, based on three years of required fuel cost data from the EHS, but not on an annual basis [13]. On the differences between indicators, TPR is an absolute indicator, while LIHC is relative. Both TPR and LIHC have their drawbacks: •TPR is a fixed ratio and very sensitive to energy prices [9]; • LIHC—smaller homes are excluded from the definition, where increases in energy prices no longer have an impact on the indicator [ 18 ]. The indicator also has a doubly-relative character, which makes it very difficult to isolate causes and effects in the process of analysis [19]. The tendency of the Hills LIHC indicator to identify low income households living in smaller dwellings was described by Moore et al. [ 20 ]. There is a disproportionate representation of households experiencing difficulty paying energy bills, and consequently under-spending on energy, resulting in lower indoor temperatures. Accounting for dwelling size is therefore critical, and a slightly different equivalization method was subsequently proposed, this method is dependent on occupancy and the dwelling size-useable floor area (m2) [20]. Dwelling size is a key factor contributing to high energy bills, therefore, in an attempt not to underestimate this effect, the adjustment of energy costs for dwelling size was deemed inappropriate by the UK Government [ 13 ]. This is coherent with other approaches, e.g., L’Observatoire National de Sustainability 2020,12, 5721 4 of 21 la Pr é carit é É nerg é tique (ONPE) statistics [ 21 ], where the effect of occupancy on energy costs is less than the impact of size of living space. Hills argues that, as energy prices change, the TPR results in thousands of households moving in and out of energy poverty [ 15 ]. In comparison, the LIHC gives stable results over time, and also across countries. This does not seem to be realistic with regard to the effect of energy prices and other variables: energy prices are comparable throughout the EU, energy efficiency is only dramatically superior in Nordic countries, with climate being the only factor which makes a significant difference regarding energy needs. To face their most severe weather, Finland, Norway, and Sweden impose higher energy efficiency standards [ 22 ], consequently heating costs are usually lower, and given that these costs are the most significant contributor to energy expenses, energy poverty indexes should be lower, as average income is higher. In Finland and Sweden, the “keep warm” subjective index is lower than 3% [ 23 ], whereas 2M retains the 15% of energy poor calculated for other European countries. In southern European countries, heating costs are high despite comparatively good weather conditions, and the “keep warm” indicator is greater than 20% in Greece and Portugal due to lower mean income, as shown in Table 1for 2M. Unexpectedly, 2M remains around 15%. The LIHC has been considered in other EU Member States as a benchmark indicator for official energy poverty statistics [ 11 ]. Rademaekers et al. [ 5 ], however, state that instead of reasonable/modeled/ theoretical energy costs, real/actual energy costs are usually used as the proxy to calculate heating expenses from official statistics. The data needed for the Hills method/LIHC indicator are provided by specially designed UK surveys; this resource-intensive and costly procedure can involve data which are difficult to obtain, making the whole process expensive. In Portugal, Spain, and Greece, for example, the current methodologies using energy poverty measurements are the following: • Greece: no official method to identify energy poverty. The “inability to keep home adequately warm” index is monitored annually by the Hellenic Statistical Authority; •Portugal: no official method to identify energy-poor households; • Spain: the only available statistics at a national level are not provided by an official institution, but thanks to the Association of Environmental Sciences (Asociaci ó n de Ciencias Ambientales, ACA) [24], although using actual bills. The EPOV performs energy poverty measurements in those three countries, and 23 more [ 25 ], based on HBS data. So, using actual energy bills. Approaches using theoretical energy costs can be found in other countries. In France, where the production of statistics by ONPE is a resource intensive process, the indicators do not reply on modeled invoices. Moore [ 18 ] found the use of actual energy bills controversial, considering this variable as a poor indicator for energy poverty. After analyzing data on actual energy use, most households do not live at the temperatures assumed in modeling [ 15 ]. In particular, the lowest income decile appear to be living at temperatures below the “adequate” threshold [ 15 ]. Moore [ 18 ] identified that low-income householders stay in cold dwellings because they spend significantly less on fuel than required. Difficulty identifying households which under-consume energetically is widespread among most Member States; according to Florio and Teissier [ 26 ], it is difficult to identify the presence of energy restrictions, indicating lower real expenditure. Palma et al. [ 27 ] also addressed this issue of under consumption through a detailed analysis of regional energy performance gaps for the city of Évora. The use of modeled energy costs, however, led researchers to include normative criteria in energy poverty characterization, resulting in the idea of factoring for “access to adequate energy services”, given that adequacy is a normative concept [ 6 ]. The justification for this approach is that modeled energy bills are related to indoor temperatures, and these temperatures are related to health problems. However, while it is difficult to know what indoor household temperature conditions are, calculating energy costs with building efficiency data is the best way to avoid this targeting problem. It is Sustainability 2020,12, 5721 5 of 21 therefore assumed that this problem does not exist in the UK, and consequently, the LIHC methodology prioritizes energy efficiency measures as the appropriate solution [28]. As an alternative to solve this problem of targeting, the Hidden Energy Poverty (HEP) index is used in Belgium [ 29 ]: households with an energy expenditure lower than 50% of the national median are also considered to be in energy poverty. In 2013, the index identified a notable 4.6% of the population as energy-poor. The application of this index highlights a new targeting problem, which can be clearly identified in the following two cases: • Highly efficient buildings cause their inhabitants (be they lowor high-income households) to be considered as energy-poor by the HEP due to the low energy costs resulting from a well-insulated building; •When comparing areas with different energy demands due to climate, those who reside in areas with a less severe climate may be considered as being in energy poverty. Thus, although the HEP addresses the problem of targeting, it causes a problem of its own. A driver in the search for alternatives is the various drawbacks associated with the Hills LIHC indicator; i.e., Moore [ 20 ] criticized its complexity and proposed several improvements, e.g., not excluding low energy costs. Notwithstanding the methodology in UK, there are few studies dealing with the use of theoretical energy costs. With the aim of addressing this gap, Imbert et al. [ 21 ] used the French 3CL-method for modeling energy costs to investigate whether it is possible, in countries where data availability is limited, to transfer energy poverty indicators based on required energy needs. In Greece, as in other European countries, the use of actual energy expenses underestimates the measure of energy poverty. Moreover, self-declared data are not reliable, as they often also include fuel consumed for transportation. To solve these problems, Papada and Kaliampakos [ 30 , 31 ] used modeled energy consumption as a variable for assessing energy poverty. There were 12 parameters that were taken into account in the Stochastic Model of Energy Poverty (SMEP) model used, e.g., heating-degree-days, coefficient of performance of heating systems, heating price, heat transfer coefficient. In Portugal, Gouveia et al. [ 4 , 32 ] showcased for all Portuguese civil parishes how a multidimensional energy poverty vulnerability index could be used, while combining information from buildings’ energy performance (i.e., energy needs), final energy consumption, climate variables, and representative socio-economic indicators. In France, based on the survey for Housing Performance, Equipment, Needs, and Uses of Energy PHEBUS (L’enqu ê te Performance de l’Habitat, É quipements, Besoins et USages de l’ é nergie), a simplified approach by income and the French energy performance label DPE (Diagnostic Performance Energ é tique) was explained by Florio and Teissier [ 26 ] for energy poverty evaluation. This approach employs the relatively simple method of setting a minimum energy performance standard for the buildings (and working with the first five deciles of income, for example), similar to the Hills alternative proposal explained above. Theoretically, if the energy performance of all homes was at maximum efficiency, and if energy prices were the same for all households, energy poverty and income poverty assessment would achieve the same results [ 10 ]. DPE, however, is only compulsory when renting or selling a property, with DPE being calculated for only 3.6 million buildings in France in 2018, and not for the remaining building stock. Non-heating expenses are not included in this method, except perhaps DHW, despite there being a strong case for revaluating the significance of non-heating energy-uses [14]. This proposal has the potential to contribute to the development of an alternative approach to the current energy poverty measures with actual energy expenses; if energy efficiency standards for buildings could be made uniform across countries, this methodology could be adapted for the national indicators of each country. Theoretical energy needs are the focus of this paper, where modeled energy bills will be calculated based on building energy performance data and non-heating energy-uses, and subsequently adjusted according to socio-demographic variables, by using equivalization weights using the Hills [ 9 ] methodology. Data from three southern EU regions, i.e., Greece, É vora municipality, Sustainability 2020,12, 5721 6 of 21 and the Basque country, will be used as case studies, with econometric analysis applied to calculate equivalization coefficients. The proposed methodology for calculating energy costs and calculations will be applied to the TPR indicator, simplifying the process to a certain extent, but not to the degree of the French DPE approach. The same approach applied to the HEP indicator facilitates the identification of households that under-consume. In order to avoid accounting for higher incomes, a cap on incomes that fall within the first five deciles can be set. This approach will avoid inaccuracies in the identification of energy-poor households. In the remaining sections, drawing on data from different European regions, this paper will focus on measuring energy poverty following the methodology above. The remainder of the article is structured as follows: in the following section the methodology will be outlined and data sets from Greece, and the Portuguese and Spanish regions will be presented. The methodology for the econometric analysis will then be described, including an explanation of why the present work is carried out using the 2M, TPR, and HEP indicators. The results of the analysis are introduced, and subsequently the key findings and discussion are presented, highlighting how the TPR and HEP indicators can improve the identification of vulnerable homes through the use of theoretical energy costs. The article concludes with some implications for policymakers. 2. Methodology In this section, details of the countries selected for the case study are provided, the methodology used for measuring energy poverty is explained, and datasets that will be tested are characterized. 2.1. Case Studies Description Three case studies were used from southern Europe (i.e., Greece, the city of É vora in Portugal, and the Basque region in Spain) to showcase the methodology, presenting interesting variations in climate characteristics, income levels, and energy prices. Case study selection was based on the presence of under-consuming and availability, in some cases even those with no heating were included. The data from Greece originated from a survey performed by Papada and Kaliampakos [ 33 ] covering the whole country. For Portugal, a door-to-door survey in the municipality of É vora was used. Information on the survey conducted, type of data collected, and key insights are available in Gouveia et al. [ 34 ]. Spanish data originated from social housing in the Basque region, accessible through the public company’s databases. Used here for the first time, it was collected in 2019, comprising data from 2018. A summary of the most relevant variables can be found in Table 1. Population data were sourced from Eurostat [ 35 ] for Greece, Portugal, Spain, and the Basque country, and PORDATA for É vora municipality [ 36 ]. The number of households was taken from Eurostat [ 35 ] (lfst_hhnhtych) for Greece, Portugal, and Spain, Eustat [ 37 ] for the Basque country, and PORDATA for É vora municipality [ 36 ]. Climate variables were heating degree days (HDD) and cooling degree days (CDD) calculated at 15.5 ◦ C. Energy prices, including taxes for households (band for electricity between 2500 and 5000 kWh/yr. and for gas between 20 and 200 GJ/yr.) were taken from the Eurostat [ 35 ]; average gas (nrg_pc_202) and electricity prices (nrg_pc_204) for the years 2016–2018. The income variable for the countries represented the adjusted gross disposable income of households per capita taken for Eurostat, while data for the city of É vora were taken from Portuguese National Statistics Office (INE) [ 38 ], and from Spanish INE for the Basque country [ 39 ]. The most relevant energy poverty indicators for Greece, Portugal, and Spain were taken from the EPOV database for latest data available [ 25 ], and referred to a percentage of the population. There were no official data available for energy poverty levels in the city of É vora, but ACA included disaggregated results for Spanish regions when calculating energy poverty indicators for Spain with data of 2016 [24]. Sustainability 2020,12, 5721 7 of 21 Table 1. Summary of most relevant variables in Greece, É vora municipality (Portugal), and Basque region (Spain). Greece Portugal City of Évora Spain Basque Region Population 2018 110,741,165 10,291,027 52,664 246,658,447 2,170,868 Number of households 2018 14,348,100 3,910,800 29,812 217,384,300 1,054,610 3 HDD (inland) 2108 41390 (Ioannina) - 1050 1500 (Madrid) 1980 (Vitoria) HDD (coast) 2018 4650 (Athens) 675 (Lisbon) - 970 (Barcelona) 955 (Bilbao) CDD (inland) 2018 41060 (Ioannina) - 1185 1355 (Madrid) 535 (Vitoria) CDD (coast) 2018 41675 (Athens) 970 (Lisbon) - 1340 (Barcelona) 875 (Bilbao) Electricity price, €per kWh, 2016–2018 10.168 0.228 0.228 50.23 0.23 5 Gas price, €per kWh, 2016–2018 10.059 0.081 0.081 50.077 0.077 5 Adjusted gross disposable income of households per capita, 2016, €1 14,622 17,686 16,644 719,216 25,121 6 High share of energy expenditure in income (2M) Population (%), 2015 8 16.3 15.1 n/a 14.2 9.0 9 Low absolute energy expenditure (M/2), Population (%), 2015 8 12.8 6.8 n/a 13.0 5.0 9 Inability to keep home adequately warm, Population (%), 2016 8 29.1 22.5 n/a 10.1 6.0 9 Arrears on utility bills, Population (%), 2016 842.2 7.3 n/a 7.8 6.0 9 1 Eurostat 2018 [ 35 ], 2 PORDATA [ 36 ], 3 Eustat [ 37 ], 4 https://www.degreedays.net/calculated at 15.5 ◦ C, 5 Market at national level, 6www.ine.es [39], 7www.ine.pt [38], 8EPOV [25], 9ACA [24]. With regard to the differences across countries, HDDs are higher in Spain, both inland and coastal areas. When comparing countries and regions, the Basque inland is colder than Madrid, and the city of É vora is considered average, as the Portuguese inland regions do not vary significantly in terms of altitude. The CDDs are greatest for the coastal areas of Greece and Spain, as the Spanish inland region is as hot as the coastal area during summer and the Greek inland area much colder. The Basque country is much colder than the rest of Spain during summer. Overall, with regard to climate data, the city of É vora is representative of the Portuguese inland regions, while the Basque region is one of the coldest in Spain. Average income is highest in Spain and lowest in Greece; the same applies to household energy prices. The disposable income in the city of É vora is only slightly below the national average. Comparatively, however, income in the Basque region is the highest. The energy poverty indicators take similar values when using relative methodologies, i.e., for 2M around 15%, but the self-perceived “inability to keep home adequately warm” is different across countries, following the inverse order of mean income. The “arrears on utility bills” indicator is disproportionately high in Greece. In the absence of official data for É vora, the only analysis on energy poverty analysis for Portuguese regions was conducted by Gouveia et al. [ 4 ]. In this study, the Energy Poverty Vulnerability Index (EPVI) was used to map and rank vulnerability to energy poverty in all Portuguese regions, both in the winter and in the summer. Their latest results revealed that the municipality of Évora is ranked 271st (for winter vulnerability) and 287th (for summer vulnerability) of a total of 308 municipalities. These results show that É vora is on the lower end of the vulnerability spectrum in the heating and cooling seasons, this can be explained by a higher adaptive capacity and lower energy performance gaps in comparison with other Portuguese regions. Finally, for the Basque country, very probably as a result of higher disposable household incomes, all the energy poverty indicators are lower. Sustainability 2020,12, 5721 8 of 21 To summarize, while the city of É vora is representative of Portugal regarding climate and income, the energy poverty levels are lower. The results of the Basque Country are not representative of the rest of Spain, due to the colder climate, higher incomes, and lower energy poverty levels. 2.2. Research Method The research hypothesis tested in the following section contests that modeled energy bills can be adjusted to actual energy expenses according to socio-demographic variables. Certain sociodemographic variables increase energy needs, for example, householders maintaining higher indoor temperatures and/or spending longer periods of time at home, the make-up of such households usually includes: (a) children under 18 years old, and (b) elderly and/or pensioners [ 13 ]. These households also very often include unemployed people and/or a dependent householder. For this reason, analysis to find correlations between actual energy bills and socio-demographic variables was first performed using data from Greece, and two regions in Portugal and Spain. The correlations between each of the variables and actual energy bills were accepted or rejected using Spearman and Pearson analyses. Subsequently, equivalization weights were calculated through an econometric analysis, so as to adjust socio-demographic variables with the actual energy expenditure and theoretical energy costs based on the available data, explained in the next section. A positive correlation between a variable and the actual energy consumption, e.g., a higher occupancy usually means higher energy consumptions, should increase the modeled energy bills. Therefore, through the econometric analysis, equivalization weights for sociodemographic characteristics were calculated. The proposed econometric model for the regression needed to calculate those equivalization factors is described in Equation (1): Log(Actual energy bills) =β1×Log(Modeled energy bills) +Σiβixi+εi, (1) where “Actual energy bills” is the dependent variable actual energy expenses, “Modeled energy bills” represents the energy costs calculated theoretically, x i are the explanatory variables (occupancy, area. and the socio-demographic variables as binary variables, namely, the presence of children and/or elderly, also including pensioners/retired and unemployed wherever possible), and εi is the error term. In order to adjust the model, βi coefficients are attained from the regression. This approach is the model commonly proposed, e.g., by [40]. Heating/cooling energy needs will be explained in the next section, and they are introduced as kWh per dwelling. Sunikka-Blank and Galvin [ 41 ] proposed using kWh per dwelling/year instead of using energy consumption per square meter (kWh/m 2 yr): they affirmed that in this way more information is given about consumers, as heating expenses depend on both the consumption per square meter and the size of the dwelling. ONPE also proposes both units to express the energy costs of the households, and for the case of LIHC/UC the equivalization considers household size and composition. However, the equivalization scale is not mentioned by ONPE [21]. Area was included as a variable in the analysis, as it is a common explanatory variable of energy consumption [ 41 ]. Occupancy is the second common variable when analyzing energy consumption, for example, through the BREDEM modeling tool, as explained in the introduction. The socio-demographic variables were binary variables, therefore multivariate regression was used. It was possible to calculate the percentage increase resulting from each of the variables, facilitating the calculation of equivalization weights for of each of the socio-demographic variables. In order not to calculate biased coefficients, since the objective was to obtain a modeled energy bill, this econometric analysis was performed for an average consumer for each of the countries, so that specific characteristics were not overrepresented, e.g., excessively low/high energy consumptions due to low/high income. In doing this, the aim of building modeled energy bills adjusted to vulnerability variables was realized. Sustainability 2020,12, 5721 9 of 21 In order to compare energy poverty levels using actual energy bills against reasonable energy costs, the obtained coefficients were incorporated in the modeled energy bills. The comparison between modeled energy bills with and without coefficients was also carried out. With regard to the indicator applied, although both the TPR and the LIHC have disadvantages, use of the former is justifiable. The introduction highlighted that, in inter-country comparison, 2M methodologies —including LIHC— do not capture the impact of energy prices on heating bills over time, other impacts, such as differences, in climate data are also omitted. The TPR was considered a good alternative given that it is intuitively understood by many. Subsequently, the analysis to test the functionality of the current methodology was carried out by calculating the TPR using actual expenses versus modeled energy bills, with and without equivalization coefficients. Additionally, this methodology for calculating reasonable energy costs can be used to identify those under consuming by comparing energy poverty measurements using actual expenses with measurements using theoretical energy needs adjusted with socio-economic variables. For Rademaekers et al. [ 5 ], under-consumption occurred when households did not spend enough to achieve a certain standard of energy services. Here, the definition assumed for under consumption was similar to the HEP indicator, this is M/2, but referred to modeled energy costs, and introducing a cap on income, this is considering only the first five deciles. If HEP underlined the existence of self-rationing practices [ 29 ], defined in HEP methodology as those consuming less than half the median (M/2), in an analogous manner, in this article it was assumed that a household was under consuming when consuming 50% of the theoretical energy cost or less. Considerations about using theoretical energy costs instead of actual expenses then applied equally to the HEP indicator. 2.3. Data Set Firstly, the data used in the correlation analysis are presented. Regarding actual expenses, occupancy, size of the dwelling, and socio-economic variables, data sources for each country/region are provided below: • For Greece, survey data from Papada and Kaliampakos [ 33 ] comprise 400 households across all of Greece, with data available on actual and required energy costs, income, building characteristics (type and size, main heating system, year of construction, region, and altitude), and some social variables (occupancy and age ranges, gender and presence of pensioners and/or unemployed). Raw data were used for the analysis performed, as recalculation of energy costs with supplementary information was unnecessary. All data were complete. The correlations between actual energy expenses and the following variables were analyzed: young under 18 years old, pensioners, people over 60, and unemployed. • In Portugal, a door-to-door survey of 388 households was performed in the municipality of É vora [ 34 ]. DHW was included as well as cooking, lighting, and electrical devices. Daily registry information from smart meters (2014–2017) was also available for this household dataset, therefore electricity consumption was retrieved for better understanding of the levels of consumption. Average electricity consumption in the sampled households was 3682 kWh [ 42 ], which represents 44% of total energy consumption, i.e., the average total energy consumption is 8350 kWh; from this total, DHW is usually non-electric and reaches 2613.5 kWh on average, approximately half of the 3081 kWh for cooking comes from electricity, electrical appliances represent 1361 kWh in electricity, and lighting an additional 560 kWh. Average data were used to fill the information gaps in actual energy expenses, in cases where no real data were available. The sample of 388 households was reduced to a final sample of 219 households, as some data from the survey were incomplete. From those 219 households, occupancy in the city of É vora was unknown for three data points and m 2 for one data point. The under 18 variable was available in this case study, as was the under 4 category, both were therefore analyzed but only the first could be compared with Greek data. There were no data available for the over 60 category, with data instead being available for Sustainability 2020,12, 5721 16 of 21 consequence their TPR and HEP indexes resulted in higher percentages of energy poverty compared to indexes calculated with equivalized modeled energy bills. Table 7. Comparison between energy poverty measurement methodologies. Index Greece Évora (Portugal) Basque Country (Spain) 2M Ratio actual bills/median 33 out of 400 (8.2%) 4 out of 145 (2.8%) 39 out of 1205 (3.2%) TPR Actual bills 190 out of 400 (47.5%) 60 out of 145 (41.4%) 428 out of 1205 (35.5%) Modeled bills 228 out of 400 (57%) 116 out of 145 (80%) 762 out of 1205 (63.2%) Modeled bills with equiv. coef. 192 out of 400 (48%) 99 out of 145 (68.3%) 715 out of 1205 (59.3%) HEP M/2 32 out of 400 (8%) 5 out of 145 (3.4%) 62 out of 1205 (5.1%) Modeled bills 64 out of 400 (16%) 112 out of 145 (77.2%) 232 out of 1205 (19.3%) Modeled bills with equiv. coef. 33 out of 400 (8.3%) 12 out of 145 (8.3%) 197 out of 1205 (16.3%) Households with no heating Total 6 out of 400 1 out of 145 17 out of 1205 Non energy-poor for 2M 6 out of 6 Yes 17 out of 17 Non energy-poor for TPR actual 3 out of 6 No 15 out of 17 Non energy-poor for TPR model 2 out of 6 No 5 out of 17 Non energy-poor for TPR model +coef 3 out of 6 No 4 out of 17 Non energy-poor for HEP median 3 out of 6 Yes 5 out of 17 Non energy-poor for HEP model 4 out of 6 No 1 out of 17 Non energy-poor for HEP model +coef 5 out of 6 Yes 2 out of 17 In the case of the TPR indexes, the obtained results when using actual bills were lower for all the cases studied. This means some actual bills are lower than their corresponding equivalized modeled bills. This was also evident when the actual bills were compared with their median (M/2) or with their corresponding equivalized modeled bills (HEP). For each case study, the following analysis will discuss if households with no heating were correctly identified by the indicators. In Greece, all the indicators had high values. This article used households with no heating as an example to compare different indexes, 2M did not identify any of these households. It is clear that the 2M indicator failed to identify them as energy-poor. Both TPR and HEP calculated with actual energy bills were complementary, with TPR identifying the high actual energy bills and HEP identifying the lower energy bills. The energy poverty indicators calculated for the city of É vora were also high. There was one household with no heating system, which should be categorized as energy-poor. Once again, 2M did not identify the household with no heating. In the Basque region, the available data for incomes were more accurate than in the previous two case studies. The indicators took high values for the 1205 households, as low-income households were overrepresented. With regard to the identification of households with no heating, there were different heating systems and different reasons for not having heating; for a multi-occupancy building with individual natural gas boilers, 6 out of 222 dwellings had no gas contract, in the other 4 buildings with prepayment central heating 6 out of the 361 dwellings had zero energy consumption, and for the remaining 8 buildings with central heating (no prepayment) 5 out of 636 dwellings had zero energy expenses. The 2M was once again the indicator with the poorest ability to identify these households. In contrast, the results of indicators with equivalized modeled bills were relatively accurate. For future work, given that two of the samples used for this study were from municipalities or regions, it might be interesting to repeat the study in order to extrapolate the results to the rest of the country/ies. Additional socio-demographic variables may also be interesting for analysis, such as long-term illness, disabilities, and hours spent at home. It will be interesting to compare temperature data, to gain insight into indoor thermal comfort levels. The issue of under-consumption is also worth investigating by research groups, i.e., in non-heating energy-uses [ 14 ], including effects on people’s quality of life. 5. Conclusions In this article the aim was to use an index that considered theoretical energy costs instead of real costs in countries/regions outside the UK. The value of developing a methodology for calculating Sustainability 2020,12, 5721 17 of 21 those theoretical energy needs combined with socio-demographic variables is that the problem of targeting households that consume less energy due to restrictions [ 26 ] and experience cold at home as a consequence is avoided [ 18 ]. The introduction highlighted that, J. Hills also introduced socio-demographic variables in his methodology, known as energy equivalization coefficients, highlighting these as important factors to consider. In order to calculate those theoretical energy costs, an econometric analysis was carried out. After comparing energy consumption levels in three southern EU region case studies—Greece, city of É vora (Portugal), and the Basque country (Spain)—the effects of certain socio-economic variables were assessed. The results are comparable to those obtained by J. Hills. These results were applied to the calculation of indexes using modeled energy bills in Portugal, Spain, and Greece. In several countries, energy poverty statistics are assessed using actual bills. Households consuming less than 50% of theoretical energy needs due to having low incomes are usually not considered as energy-poor according to the TPR using actual expenses. This is a common problem for most European countries. The HEP was developed to identify these households, but when using the median of the actual bills, this index inaccurately captures low energy costs due to high building energy efficiency. This can be avoided using reasonable energy costs, as in Hills [ 9 ], or modeled required energy consumption, as in Papada and Kaliampakos [ 30 ], the methodology used herein. The conclusion reached for current energy poverty indicators, in line with the opinion of researchers referenced in the introduction such as Moore [ 18 ] and Florio and Teissier [ 26 ], among others, is that actual expenses is not a good proxy for energy poverty measurement; the fact that under-consuming households and even those with no heating are not considered as energy-poor leads to a targeting problem. Therefore, theoretical energy costs should be used in energy poverty indexes, including in those modeled bills the equivalization factors, as proposed by Hills [ 9 ]. This is also effective in inter-country comparison, as theoretical expenses can be excessively high in countries with low mean incomes, such as those in the southern EU regions studied. Regarding comparison across countries, building an index that does not reflect the impact of energy over time or calculates similar percentages of energy-poor households across countries appears unrealistic. This is the case of indicators based on 2M. A pivotal role should be played by the energy costs-to-income ratio; this means that trade-offs can occur in homes trying to keep their houses warm, e.g., cutting food expenses, or they can fall into debt. This is the so called “heat or eat” dilemma when energy costs in fuel-poor households compete with other basic expenses, which can lead to health consequences [ 15 ]. TPR may seem to be arbitrary, setting the amount of energy expenses at 10%, or a normative way of sharing the income among the different expenses. Evaluating energy needs should be performed for the desired outcome, calculating the share of the income destined to energy bills across European countries through the energy costs-to-income ratio approach. For policymakers, rather than energy prices and energy efficiency, the important variable is energy costs, namely theoretical energy costs. With actual bills, the index does not capture the whole picture. With modeled energy costs, information on both energy efficiency and energy prices is captured: fuel poverty is distinct from income poverty [ 9 ], being caused primarily by poor energy efficiency and availability of affordable energy carriers. In the end, the focus should be placed on favoring energy efficiency measures. This also has implications for carbon reduction: in the case of households under-using energy, taking into account only actual bills, investments to reduce carbon emissions cannot be recovered. Other types of benefits—health, employment, indoor air quality, and thermal comfort—should be considered to make those energy efficiency measures ‘profitable’ [51]. A methodology like this has also implications for future legislation or official standards [ 6 ]. The data extracted from HBS are those related to the actual energy bills. If modeled energy bills are to be used, the problem is that the existing building standards for energy efficiency are different across EU countries. The Belgium approach, where the overall heat transfer coefficient of the building envelope (U-value) was advocated by Sunikka-Blank and Galvin [ 41 ] for this purpose, this approach is also comparable to the German EPR. However, increasing research on the role of EPCs in energy poverty Sustainability 2020,12, 5721 18 of 21 assessment and retrofitting of buildings (e.g., Gouveia and Palma [ 32 ] or Charalambides et al. [ 52 ]), as well as new ways of accessing European building characteristic data are now available, as the EU Commission launched an initiative to collect data on energy efficiency standards from building stock [ 53 ]. The EU SILC data can be used to adjust those socio-economics variables in all EU countries, as well as the Household Budget Survey (HBS). Author Contributions: Conceptualization, I.A.; methodology, L.P. and I.A.; data access, L.P., J.P.G. and I.A.; writing—original draft preparation, I.A.; writing—review and editing, L.P. and J.P.G.; Short Term Scientific Mission administration, N.K. and D.K. All authors have read and agreed to the published version of the manuscript. Funding: Iñigo Antepara thanks the ENGAGER Action CA16232 “European Energy Poverty: Agenda Co-Creation and Knowledge Innovation” for the Short Term Scientific Mission scholarship awarded that was completed during March–April 2018 at NTUA, Athens (Greece). Jo ã o Pedro Gouveia acknowledge and thank the support given to CENSE by the Portuguese Foundation for Science and Technology (FCT) through the strategic project UIDB/04085/2020. The paper stems from collaborative work within COST Action ‘European Energy Poverty: Agenda Co-Creation and Knowledge Innovation’ (ENGAGER 2017–2021, CA16232) funded by European Cooperation in Science and Technology—www.cost.eu. Acknowledgments: The authors would like to thank Katherine Mahoney for the English language review. Iñigo Antepara would like to thank Alokabide for allowing access to data. Conflicts of Interest: The authors declare no conflict of interest. Appendix A Table A1. Correlation analysis for Greek data. Variable Number of Observations Spearman’s rho Pearson Actual energy expenses (log) vs occupancy 400 0.2964 ** 0.3028 ** Actual energy expenses (log) vs m2400 0.4577 ** 0.47 ** Socio-demographic variables Actual energy expenses (log) vs u18 yr. old 400 0.1386 ** 0.1381 ** Actual energy expenses (log) vs Pensioners 400 −0.1554 ** −0.1599 ** Actual energy expenses (log) vs over 60 400 −0.0873 ns −0.0822 ns Actual energy expenses (log) vs Unemployed 400 −0.0866 ns −0.1019 * * significant at the 0.05 level, ** significant at the 0.01 level, and ns not statistically significant. Table A2. Correlation analysis for Portuguese data. Variable Number of Observations Spearman’s rho Pearson Actual energy expenses (log) vs occupancy 216 0.4047 ** 0.3744 ** Actual energy expenses (log) vs m2218 0.3419 ** 0.2666 ** Socio-demographic variables Actual energy expenses (log) vs u4 yr. old 219 0.1135 ns 0.0686 ns Actual energy expenses (log) vs u18 yr. old 219 0.2531 ** 0.2323 ** Actual energy expenses (log) vs Retired 219 −0.1763 ** −0.1537 * Actual energy expenses (log) vs over 65 219 −0.1749 ** −0.1476 * * significant at the 0.05 level, ** significant at the 0.01 level, and ns not statistically significant. Table A3. Correlation analysis for Basque data. Variable Number of Observations Spearman’s rho Pearson Actual energy expenses (log) vs occupancy 1216 0.1137 ** 0.1097 ** Actual energy expenses (log) vs m21219 0.0389 ns 0.0454 ns Socio-demographic variables Actual energy expenses (log) vs u4 yr. old 489 −0.1215 ** −0.12 ** Actual energy expenses (log) vs u18 yr. old 489 −0.1376 ** −0.1234 ** Actual energy expenses (log) vs over 65 505 −0.0323 ns −0.0629 ns ** significant at the 0.01 level, and ns not statistically significant. Sustainability 2020,12, 5721 19 of 21 Appendix B Table A4. Equivalization factors proposed by Hills [9]. Household Type Equivalization Factor Couple with dependent children 1.15 Couple without dependent children 1.00 Lone parent 0.94 Single person 0.82 Other multi-person household 1.07 Table A5. Proposed energy cost equivalization factors by Moore et al. [20]. Household Size Dwelling Size-Useable Floor Area (m2) <50 50–69 70–89 90–109 110+ Single person 0.63 0.77 0.90 0.96 1.22 Two people 0.70 0.84 1.00 1.08 1.28 Three people 0.78 0.86 1.01 1.17 1.44 Four people 0.84 0.92 1.07 1.20 1.53 Five+people 0.87 1.00 1.14 1.31 1.62 References 1. Database–Eurostat. Available online: https://ec.europa.eu/eurostat/data/database (accessed on 29 March 2020). 2. Day, R.; Walker, G.; Simcock, N. Conceptualising energy use and energy poverty using a capabilities framework. Energy Policy 2016,93, 255–264. [CrossRef] 3. European Commission EU Energy Poverty Observatory 2018. Available online: https://www.energypoverty.eu/ (accessed on 21 June 2020). 4. Gouveia, J.P.; Palma, P.; Simoes, S.G. Energy poverty vulnerability index: A multidimensional tool to identify hotspots for local action. Energy Rep. 2019,5, 187–201. [CrossRef] 5. Rademaekers, K.; Yearwood, J.; Ferreira, A.; Pye, S.; Hamilton, I.; Agnolucci, P.; Grover, D.; Kar á sek, J.; Anisimova, N. Selecting Indicators to Measure Energy Poverty; Trinomics: Rotterdam, The Netherlands, 2016. 6. Schuessler, R. Energy Poverty Indicators: Conceptual Issues-Part I: The Ten-Percent-Rule and Double Median/ Mean Indicators; Discussion Paper 14-037; ZEW-Centre for European Economic Research: Mannheim, Germany, 2014. 7. Boardman, B. Fuel Poverty: From Cold Homes to Affordable Warmth; Belhaven Press: London, NY, USA, 1991; ISBN 1-85293-139-6. 8. DECC. Fuel Poverty Methodology Handbook; BRE: London, UK, 2010. 9. Hills, J. Getting the Measure of Fuel Poverty: Final Report of the Fuel Poverty Review; Centre for Analysis of Social Exclusion; The London School of Economics and Political Science; Commissioned by Department of Energy and Climate Change (DECC): London, UK, 2012. 10. Probert, L.; Haines, V.; Loveday, D. The Hills Fuel Poverty Review Interim Report: Assessing Proposals and Implications. People Place Policy Online 2011,5, 161–173. [CrossRef] 11. Sareen, S.; Thomson, H.; Herrero, S.T.; Gouveia, J.P.; Lippert, I.; Lis, A. European energy poverty metrics: Scales, prospects and limits. Glob. Transit. 2020,2, 26–36. [CrossRef] 12. Liddell, C.; Morris, C. Fuel poverty and human health: A review of recent evidence. Energy Policy 2010,38, 2987–2997. [CrossRef] 13. DECC. Fuel Poverty Methodology Handbook; BRE: London, UK, 2016. 14. Simcock, N.; Walker, G.; Day, R. Fuel poverty in the UK: Beyond heating. People Place Policy 2016 ,10, 25–41. [CrossRef] 15. Hills, J. Fuel Poverty: The Problem and Its Measurement. Interim Report of the Fuel Poverty Review; Centre for Analysis of Social Exclusion, LSE: London, UK, 2011. Sustainability 2020,12, 5721 20 of 21 16. Hagenaars, A.J.; De Vos, K.; Asghar Zaidi, M. Poverty Statistics in the Late 1980s: Research Based on Micro-Data; Office for Official Publications of the European Communities: Luxembourg, 1994. 17. Bouzarovski, S.; Petrova, S. A global perspective on domestic energy deprivation: Overcoming the energy poverty–fuel poverty binary. Energy Res. Soc. Sci. 2015,10, 31–40. [CrossRef] 18. Moore, R. Definitions of fuel poverty: Implications for policy. Energy Policy 2012,49, 19–26. [CrossRef] 19. Heindl, P.; Schuessler, R. Dynamic properties of energy affordability measures. Energy Policy 2015 ,86, 123–132. [CrossRef] 20. Moore, R.; Bridgeman, T.; Croft, D.; Guertler, P.; Preston, I. Improving the Hills Approach to Measuring Fuel Poverty; Centre for Sustainable Energy: London, UK, 2012. 21. Imbert, I.; Nogues, P.; Sevenet, M. Same but different: On the applicability of fuel poverty indicators across countries—Insights from France. Energy Res. Soc. Sci. 2016,15, 75–85. [CrossRef] 22. Healy, J.D. Excess winter mortality in Europe: A cross country analysis identifying key risk factors. J. Epidemiol. Community Health 2003,57, 784–789. [CrossRef] [PubMed] 23. Bouzarovski, S. ENGAGER Policy Brief 1, COST Action CA16232; University of Manchester: Manchester, UK, 2018. 24. De Ciencias Ambientales, A. Pobreza Energ é tica en España: Hacia un Sistema de Indicadores y una Estrategia de Actuaci ó n Estatal [Energy Poverty in Spain: Towards a System of Indicators and a State Action Strategy]; Asociaci ó n de Ciencias Ambientales: Madrid, Spain, 2018. 25. EU Energy Poverty Observatory Indicators & Data. Available online: https://www.energypoverty.eu/ indicators-data (accessed on 6 June 2020). 26. Florio, P.; Teissier, O. Estimation of the Energy Performance Certificate of a housing stock characterised via qualitative variables through a typology-based approach model: A fuel poverty evaluation tool. Energy Build. 2015,89, 39–48. [CrossRef] 27. Palma, P.; Gouveia, J.P.; Simoes, S.G. Mapping the energy performance gap of dwelling stock at high-resolution scale: Implications for thermal comfort in Portuguese households. Energy Build. 2019 ,190, 246–261. [CrossRef] 28. Middlemiss, L. A critical analysis of the new politics of fuel poverty in England. Crit. Soc. Policy 2017 ,37, 425–443. [CrossRef] 29. Meyer, S.; Laurence, H.; Bart, D.; Middlemiss, L.; Mar é chal, K. Capturing the multifaceted nature of energy poverty: Lessons from Belgium. Energy Res. Soc. Sci. 2018,40, 273–283. [CrossRef] 30. Papada, L.; Kaliampakos, D. A Stochastic Model for energy poverty analysis. Energy Policy 2018 ,116, 153–164. [CrossRef] 31. Papada, L.; Kaliampakos, D. Development of vulnerability index for energy poverty. Energy Build. 2019 ,183, 761–771. [CrossRef] 32. Gouveia, J.P.; Palma, P. Harvesting big data from residential building energy performance certificates: Retrofitting and climate change mitigation insights at a regional scale. Environ. Res. Lett. 2019 ,14, 095007. [CrossRef] 33. Papada, L.; Kaliampakos, D. Measuring energy poverty in Greece. Energy Policy 2016 ,94, 157–165. [CrossRef] 34. Gouveia, J.P.; Seixas, J.; Long, G. Mining households’ energy data to disclose fuel poverty: Lessons for Southern Europe. J. Clean. Prod. 2018,178, 534–550. [CrossRef] 35. European Commission Eurostat. Available online: https://ec.europa.eu/eurostat/(accessed on 16 November 2019). 36. PORDATA PORDATA—Estat í sticas, gr á ficos e Indicadores de Munic í pios, Portugal e Europa. Available online: https://www.pordata.pt/(accessed on 16 November 2019). 37. EUSTAT Datos estad í sticos de la C.A. de Euskadi. Available online: http://es.eustat.eus/indice.html (accessed on 17 November 2019). 38. INE Portal do INE. Available online: https://www.ine.pt/xportal/xmain?xpgid=ine_main&xpid=INE (accessed on 6 June 2020). 39. INE. Instituto Nacional de Estadística. Available online: https://www.ine.es/(accessed on 6 June 2020). 40. Bair, S.; Belaid, F.; Teissier, O. Quels enseignements tirer de l’enquete Phebus sur la question de l’effet rebond? Les Ménages et La Consomm. d’Énergie 2017, 101–113. 41. Sunikka-Blank, M.; Galvin, R. Introducing the prebound effect: The gap between performance and actual energy consumption. Build. Res. Inf. 2012,40, 260–273. [CrossRef] 42. INE. Consumo de Energia no Sector Doméstico 2010; INE; DGEG: Lisboa, Portugal, 2011. 43. De Ministerio, V. C ó digo T é cnico de la Edificaci ó n. Documento B á sico HE; Ahorro de energ í a: Madrid, Spain, 2017. Sustainability 2020,12, 5721 21 of 21 44. EVE. Claves energéticas del sector doméstico en Euskadi; Ente Vasco de la Energía: Bilbao, Spain, 2013. 45. University of Oxford Our World in Data 2019. Available online: https://ourworldindata.org/(accessed on 21 June 2020). 46. Raynaud, M.; Bourges, B.; Osso, D.; Duplessis, B.; Adnot, J. Bottom-up Statistical Analysis of the Energy Consumption of French Single-Family Dwellings; hal-01133743; Archive ouverte HAL: Lyon, France, 2014. 47. Denjean, M. Comportements et consommations é nerg é tiques. Les M é nages et La Consomm. d’ É nergie 2017 , 60–67. 48. Cavailh è s, J.; Joly, D.; Brossard, T.; Cardot, H.; Hilal, M.; Wavresky, P. La Consommation d’ É nergie des M é nages en France; INRA/CNRS for the CGDD: Dijon, France, 2011. 49. Risch, A.; Salmon, C. What Matters in Residential Energy Consumption? Evidence from France; hal-01081953; Archive Ouverte HAL: Lyon, France, 2013. 50. Tirado Herrero, S. Energy poverty indicators: A critical review of methods. Indoor Built Environ. 2017 ,26, 1018–1031. [CrossRef] 51. Ugarte, S.; van der REE, B.; Voogt, M.; Eichhammer, W.; Ordoñez, J.A.; Reuter, M.; Schlomann, B.; Lloret Gallego, P.; Villafafila Robles, R. Energy Efficiency for Low-Income Households; Policy Department A: Economic and Scientific Policy; European Parliament: Brussels, Belgium, 2016. 52. Charalambides, A.G.; Maxoulis, C.N.; Kyriacou, O.; Blakeley, E.; Frances, L.S. The impact of Energy Performance Certificates on building deep energy renovation targets. Int. J. Sustain. Energy 2019 ,38, 1–12. [CrossRef] 53. European Commission EU Building Stock Observatory. Available online: https://ec.europa.eu/energy/topics/ energy-efficiency/energy-efficient-buildings/eu-bso_en (accessed on 3 June 2020). © 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).