Measuring the thermal energy performance gap of labelled residential buildings in Switzerland

Energy Policy xxx (xxxx) xxx

Contents lists available at ScienceDirect

Energy Policy journal homepage: http://www.elsevier.com/locate/enpol

Measuring the thermal energy performance gap of labelled residential buildings in Switzerland Stefano Cozza *, Jonathan Chambers, Martin K. Patel Chair for Energy Efficiency, Institute for Environmental Sciences and Department F.-A. Forel for Environmental and Aquatic Sciences, University of Geneva, Switzerland

A R T I C L E I N F O

A B S T R A C T

Keywords: Energy performance gap Energy label Actual consumption Environmental policy target

This paper addresses the thermal Energy Performance Gap (EPG), defined as the difference between a building’s theoretical and actual energy consumption for thermal purposes (heating and hot water). Successful energy policies require estimates of the energy saving potential of the building stock. It is the objective of this work to analyse whether and to what extent an EPG exists in residential buildings in Switzerland. The database of the Swiss Cantonal Energy Certificate for Buildings was used, covering over 50 000 buildings. The median EPG was found to be 11% (i.e. actual consumption lower than theoretical) but varied across ratings from 12.4% (B-label) to 40.4% (G-label). Buildings with low energy ratings tend to consume signifi­ cantly less than expected, while buildings with high rating tend to consume slightly more than expected. For the A-labels buildings (0.5% of the total) an EPG of 6.2% was found, suggesting that the very high-performance buildings may be more robust to the EPG. Simplified scenarios to illustrate the impact of this EPG on total consumption are presented, which highlight the challenge of meeting the Swiss Energy Strategy 2050 with a realistic renovation rate. The importance of low carbon heat supply for buildings is also discussed.

1. Introduction Buildings are currently responsible for 40% of energy consumption and 36% of CO2 emissions in the European Union (EU) (Arcipowska et al., 2016). Therefore they play a major role towards achieving the energy efficiency targets set in the EU (EU Parliament, 2014). Parallel to the development of these new objectives in EU energy policy, the Swiss Federal Council has been developing and implementing the Swiss En­ ergy Strategy 2050 (ES-2050) since 2011 (Swiss Federal Office of En­ ergy, 2018). The ES-2050 is based on three strategic objectives: increasing energy efficiency, increasing the use of renewable energy, and withdrawal from nuclear energy. For residential buildings, this translates into a target of 29 TWh/y consumed and 2.7 Mt CO2/y emitted by 2050, with the final energy consumption to be reduced by 46% and CO2 emissions by 77% compared to today levels according to the so-called “New Energy Policy” (Prognos, 2012). Buildings are a key focus area of European energy and climate policy. With the first version of the EU’s Energy Performance of Buildings Directive (EPBD), an initial step to regulate the energy consumption in buildings at EU level was taken in 2002 (The European Parliament and

the Council of the EU, 2003). Aside from several measures targeted to reduce CO2 emissions, energy performance certificates were introduced under the EPBD, for buildings that are constructed, sold or rented out to a new tenant, to raise awareness, influence the decision of buyers, owners and tenants and to ensure that performance fulfils the minimum national requirement (Sutherland et al., 2015). Since then, the revised EPBD (The European Parliament and the Council of the EU, 2018) which amends parts of the 2010 EPBD, was introduced in 2018, confirming pre-existing objectives and adding some new elements. For example, all new buildings must be nearly zero-energy buildings from 31 December 2020, all EU countries will have to express their national energy per­ formance requirements in ways that allow cross-national comparisons, and set cost-optimal minimum energy performance requirements for new buildings and for the renovation of existing ones. For energy per­ formance certificates, the EPBD establishes the requirements for stan­ dardized calculation procedures, which may differ from country to country but are commonly based on a calculation of the expected energy use (Petersen and Hviid, 2012). However, it can be questioned whether the theoretical energy use reported in this kind of certificates is really achieved in practice, i.e. whether buildings are actually using the expected amount of energy.

* Corresponding author. E-mail address: [email protected] (S. Cozza). https://doi.org/10.1016/j.enpol.2019.111085 Received 13 February 2019; Received in revised form 18 October 2019; Accepted 1 November 2019 0301-4215/© 2019 Elsevier Ltd. All rights reserved.

Please cite this article as: Stefano Cozza, Energy Policy, https://doi.org/10.1016/j.enpol.2019.111085

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

(2011), the Standard Assessment Procedure to evaluate the building efficiency in UK is used to calculate theoretical consumptions. Also in this case, it was found that buildings with better performance tend to consume more than calculated, while worse-performing buildings tend to consume less. Guerra-Santin and Tweed (2015) discussed the many factors that give rise to uncertainty in the theoretical energy consumption. The three main inaccuracies that have been identified in the literature are the expected indoor air temperature (Gaetani et al., 2016; Hughes et al., 2015; Khoury et al., 2017; Zhang et al., 2018), the U-values assumed for building façade elements (Ahern and Norton, 2019; Hoffmann and Geissler, 2017; Hughes et al., 2015; Lehmann et al., 2017; Loucari et al., 2016), and the expected air change rate (Burman et al., 2014; Delghust et al., 2015; Grossmann et al., 2016; Kragh et al., 2017; La Fleur et al., 2017). According to recent literature, these factors, that are usually set by standards, are generally overestimated for inefficient building (e.g. actual indoor temperature below the standard) and underestimated for efficient buildings (e.g. actual indoor temperature above the standard), leading to a different EPG (Khoury et al., 2018; Tian et al., 2018; Van Dronkelaar et al., 2016; Zou et al., 2018). Multiple analyses have, however, revealed that the implementation of energy performance certificates differs across countries, and consequently the findings from previous studies cannot be assumed to apply to any other country (Andaloro et al., 2010; Delghust et al., 2015). The overconsumption of high performing buildings has been noted in several case studies in Switzerland, where the actual consumption of new residential buildings was found to be larger than expected, with an EPG of þ42% for 78 buildings (Reimann et al., 2016), þ55% for 3 buildings (Thaler and Kellenberger, 2017), þ70% for 10 buildings (Zgraggen, 2010). In all these studies the theoretical consumption are obtained following the Swiss norms (SIA, 2015) and the actual con­ sumptions are obtained through in-situ measurements. A difference in the actual consumption relative to expectations can invalidate a significant part of the calculated energy savings. For example, the use of the theoretical rating to evaluate the energy savings that can be achieved through energy retrofit may misrepresent the actual savings, invalidating the assumptions of future scenarios. Research for Germany found that, even by retrofitting all homes to the highest standard, only half of the expected savings would be achieved, due to the existing EPG (Sunikka-Blank and Galvin, 2012). Therefore, the existence of the EPG can be critical for the success of energy effi­ ciency policies. It is currently unclear whether and to what extent these findings can be applied to the Swiss building stock, as research so far has focused on a few case studies or small samples (Hoffmann and Geissler, 2017; Leh­ €ssig, mann et al., 2017; Thaler and Kellenberger, 2017; Wyss and Ha 2016). The largest Swiss research study analysed 214 buildings (Reim­ ann et al., 2016), including only newly built dwellings and offices, and is therefore not representative of the entire building stock. Furthermore, a report recently published by the Swiss Federal Office of Energy states: “from literature it cannot be inferred whether an EPG exists in the Swiss building park as a whole. The main reasons are the small database and an imprecise handling of definitions and key assumptions” (Frei et al., 2018). Finally, given that the Gross Domestic Product per capita of Switzerland (in purchasing power parity terms, as indicator of a coun­ try’s standard of living) is almost twice that of the EU average (65 vs. 37 thousand-$; World Bank, 2018), it can be questioned whether in­ habitants of buildings with poor energy performance show a similar behaviour as observed in other European countries, whether the overall thermal performance of the Swiss building stock differs, and what effect this has on the EPG (Schuler et al., 2000). The overarching research question of this paper is therefore to establish whether there is an energy performance gap in residential buildings in Switzerland. If so, this leads to the subsequent questions of how the EPG is distributed among the different building typologies (energy label and age) and what this may imply for the achievement of

Acronyms CECB EPBD EPG ERA ES-2050 EU FSO MFH NEP SFH SIA

Swiss Cantonal Energy Certificate for Buildings Energy Performance of Buildings Directive Energy Performance Gap energy reference area Swiss Energy Strategy 2050 European Union Swiss Federal Statistical Office multi-family house New Energy Policy Single family house Swiss Society of Engineers and Architects

Wherever this is not the case one speaks of the Energy Performance Gap (EPG), defined as the difference between measured and calculated en­ ergy consumption of a building (de Wilde, 2014; Sunikka-Blank and Galvin, 2012). Although there are various approaches to analyse the EPG, the definition of the difference in consumption as a proportion of the theoretical consumption for a given building is broadly accepted (Galvin, 2013): EPG ½%� ¼

Actual consumption Theoretical consumption Theoretical consumption

(1)

Several European projects were launched to monitor the actual performance of buildings and reduce the EPG, such as EPISCOPE (Das­ calaki et al., 2016), which aims to track the progress towards targeted savings in several European countries, or TRIME (Meijer, 2017), that studies the performance gap through monitoring the behavioural change of the occupants after retrofits. For newly constructed buildings, sig­ nificant efforts have been put into the operative phase (facility man­ agement, building management), with the development of new tools for minimising the EPG (such as smart building controls), through projects such as TRIBUTE (Pietropaoli et al., 2014) and HIT2GAP (Andri�c et al., 2017). Moreover, a range of studies have been published in recent years focusing on the EPG and its consequences for meeting the national en­ ergy saving goals across Europe (Calì et al., 2016; Delghust et al., 2015; Loucari et al., 2016; Merzkirch et al., 2014). Although it is difficult to establish direct relations between the magnitude of the EPG and the local national standards, a common pattern has been found on how the EPG depends on a building’s thermal performance: there is a broad agreement in the literature that buildings with poor thermal performance (low energy rating) tend to consume less than predicted. Vice versa, buildings with high thermal performance (high energy rating) tend to consume more than predicted (Merzkirch et al., 2014; Ramallo-Gonz� alez, 2013; Risholt and Berker, 2013; Sharpe and Shearer, 2013). For example, Majcen (2016) found that very effi­ cient buildings in the Netherlands have a positive EPG (þ20%, consuming more than expected) while very inefficient buildings have a negative EPG ( 50%, i.e. actual consumption was half the theoretical consumption). Cayre et al. (2011) arrived at similar finding for France using the performance certificates provided by the national database for 923 residential buildings. Results show that calculations based on the standard model strongly overestimate space heating consumption in older housing as well as their energy savings potential. Likewise for Belgium, Hens et al. (2010) used the methodology imposed by the en­ ergy performance regulation to calculate the consumption of 964 old houses to conclude that measured energy consumption for heating res­ idential buildings may be only half the calculated value when assuming a standard use. In another research, Sunikka-Blank and Galvin (2012) examined the German building stock dwellings using the national norms to evaluate the performance of 3400 houses. The results indicate that actual consumption is, on average, 30% less than calculated. In Kelly 2

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

the ES-2050 objectives. Given their importance of space heating and domestic hot water for the thermal performance, this study considers both of these components when studying the EPG (while excluding electricity use for appliances, lighting and any other purpose). This study, covering 50 000 building energy certificates, is the most comprehensive so far conducted in Switzerland and comparable to the most extensive research in Europe (200 000 buildings in the Netherlands €der et al., (Majcen et al., 2013) and 140 000 buildings in Germany (Schro 2014)). This work responds to the stated need for a general quantifica­ tion of the EPG taking into account a large sample of buildings and their respective theoretical and actual consumption, as addressed in various international (de Wilde, 2014; Galvin, 2013; Sunikka-Blank and Galvin, 2012) and Swiss (Frei et al., 2018; Khoury et al., 2016) studies. It pro­ vides statistically significant results on the residential buildings’ per­ formance which can offer valuable new insights for policy makers, energy utilities, local authorities, building owners and researchers. Moreover, this paper builds on studies published for several EU coun­ tries, including Belgium, the Netherlands, Austria, and Germany, by contributing results for Switzerland. This paper is divided into six sections. Section 2 gives a brief over­ view of the Swiss energy label system for buildings. Section 3 presents the method and the dataset used for the analysis. The results for different scenarios are presented and discussed in sections 4 and 5. Finally, in section 6, conclusions are drawn together with their policy implications.

The actual consumption obtained in this manner is used in the CECB to check the building model adopted for the theoretical consumption. These actual consumption data allow to identify the effect of different user behaviour or malfunctioning of the heating system. Therefore, both theoretical and actual energy consumption are essential for studying the EPG. 3. Method and data 3.1. Method To calculate the EPG, Equation 1 is applied by using the values for the actual and theoretical energy consumption from the CECB described above. The performance gap analysed in this work is based on the dif­ ference between the real energy consumption and the theoretical energy consumption according to regulations assuming normalised operating conditions as set out in the national standards (Van Dronkelaar et al., 2016). In order to compare the actual consumption on an equal basis, all the different energy carriers used (e.g. m3 of gas, litres of oil, kg of biomass) have been converted to kWh/(m2y) of final energy using the standard conversion factors given by the Swiss norms (Appendix B; SIA, 2016). Therefore, this work only considers theoretical consumption for thermal use (space heating and domestic hot water) as final energy in kWh/(m2y), which is directly comparable to the actual final energy consumption for thermal use provided by energy bills (which is reported in the dataset as a separate value from the gross total energy con­ sumption). The distributions of EPG values and energy consumption values as a function of the different energy ratings were analysed, in order to unveil trends or patterns in the EPG as a function of the building’s performance. To establish the difference between actual and calculated emissions, the CO2 emissions of the CECB were calculated using the KBOB energy emissions intensity database, based on the Ecoinvent methodology (KBOB, 2016). This database assesses the greenhouse effect as the cu­ mulative effects of the different greenhouse gases and express it as kg CO2-eq. This value makes it possible to associate the heat source of each building in the CECB sample with the emissions factor of the KBOB database, and therefore evaluate the total CO2 emissions.

2. The Swiss energy label for buildings Following a similar rationale as the EPBD, Switzerland developed its own standard for the energy certification for buildings (SIA, 2031). Based on the European standards SN EN 15217 and SN EN 15603, the Swiss Cantonal Energy Certificate for Buildings (CECB) was created in 2009. It reports the energy efficiency of the building envelope and the energy requirements if the building performs in compliance with the standards. This applies both to existing and new buildings. The calcu­ lated performance is categorized into the energy labels A to G (very efficient to very inefficient). Currently, the CECB is not mandatory everywhere in Switzerland, as it was initially proposed as a voluntary measure (Conferenza Cantonale dei Direttori dell’Energia CDE, 2014). However, an increasing number of cantons are making it compulsory to issue this certificate when real estate is for sale, in order to be eligible for subsidies for refurbishment measures, or for new buildings. The CECB delivers two ratings, one considers only the building en­ velope (walls, roof, windows, etc) while the second considers the overall energy efficiency including heating system, domestic hot water, and other loads (e.g. appliances). The rating is performed based on primary energy demand. The energy sources used are weighted using the na­ tional energy factors (EnDK, 2016): the use of renewable energy and/or a heat pump contributes to a better classification. This study will only consider the data associated with overall energy efficiency rating, which includes the breakdown of final energy consumption for space heating and domestic hot water. The theoretical consumption represents the energy use of the building under the standard conditions of occupation (Appendix A) and weather (SIA, 2028). The heat balance calculations follow SIA 380, based on the static monthly balance indicated in the European SN EN 13790. This implies that variations due to user behaviour and outdoor weather conditions are disregarded. The methodology to measure the actual energy consumption is codified in standard SIA 2031. Actual energy use is determined as an average of the measurements taken over at least three consecutive years. If the annual consumption in a particular year presents a variation of more than 20% of the average of the other measured years, the measures of that year shall not be taken into account. These data are usually ob­ tained from energy invoices for the different energy carriers used in the building: fossil fuels or electricity for the heating system and electricity for lighting, ventilation system and appliances.

3.2. CECB dataset The CECB dataset includes general building metadata (location, closest climate station, construction year), geometry (dimensions, en­ ergy reference area (ERA), orientation), envelope quality information (areas, U-vales of each element) as well as type of heating system (fossil, heat pump, solar thermal). Given the diversity of information contained in the dataset and their heterogeneity in terms of level of detail and accuracy of the inspection, a pre-processing of the data was needed before proceeding with any analysis. There were 51 318 residential buildings (both single family houses/SFH and multi-family houses/ MFH) in the original dataset. Any certificate with corrupted information was deleted, e.g. impossibly small heated area (<1 m2) or negative Uvalue for the envelope. Since some certificates had undergone several revisions, with all revisions stored in the dataset, only the latest version was kept as in most cases the previous versions were temporary incomplete copies created by the CECB Expert. After these operations the dataset was reduced to 43 639 buildings (85% of the initial dataset). Similar filtering operations with comparable error ratio were found in several publications dealing with national energy data, as the SHAERE database (Social Rental Sector Audit and Evaluation of Energy Saving Results) in van den Brom et al. (2017), the VEA database (Flemish Energy Agency) in Delghust et al. (2015) and the CBS database (Statistics Netherlands) used by Majcen et al. (2013). Only the certificates with complete information on actual and theoretical consumption were retained, as in many cases the actual 3

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

consumption was missing due to the lack of three years of energy bills. This requirement reduced the sample to 36 299 buildings. Finally, the outliers in the energy consumption were removed. Outliers in this case are buildings with energy consumption values which are extremely different from other buildings and which were hence considered to be highly atypical buildings or possibly erroneous data. To identify them, the modified Z-scores method was used as proposed by Iglewicz and Hoaglin (1993) with a score threshold of 3.5 as suggested by the authors. This last filtering operation reduced the sample to the final size of 34 816 residential buildings (68% of the initial dataset). 3.3. Representativeness of the sample An important advantage of using this dataset is the number of buildings and its geographical coverage of Switzerland as a whole. Nevertheless, after the operations of cleaning and filtering of the data, it was important to test whether the resulting sample was representative of the Swiss buildings stock. The national register of buildings, published by the Swiss Federal Statics Office (FSO, 2017) was used as a point of comparison in order to check the representativeness of the CECB sample. According to this register, at the end of 2017, there were 2 million buildings in total In Switzerland of which 1.7 million were residential. Of these, 57% were SFH and 26% MFH (FSO, 2017). The residential buildings account for 32% of Switzerland’s final energy consumption, representing a total of 67 TWh/y including space heating, cooling, ventilation, domestic hot water, lighting, and general electricity con­ sumption. The largest share of this energy consumption is devoted to space heating and domestic hot water, i.e. 55 TWh/y (Prognos, 2017), representing the focus of this study. Fig. 1 compares the heating system used in the buildings covered in the CECB sample, revealing an excellent match with the building stock. The small mismatch for direct electric heating (Electric) and heat pumps (HP) can be partially explained with an incorrect identification of the heating system by the CECB Experts (meaning by Electric the energy source and not the heating system), as discovered later in the analysis. In Fig. 2 the building construction period is compared, showing overrepresentation of the buildings constructed between 1945 and 1990. This can be partially explained by the original focus of the CECB on buildings in need of refurbishment (but which was less applicable to historic buildings pre-1945). We can only speculate as to the reasons behind the other discrepancies shown in Fig. 2. The underrepresentation in the CECB sample between the 1990 and 2010 may be explained by the simultaneous introduction of the Minergie standard at the end of the 1990s (Minergie, 2010). The Minergie standard is a Swiss quality label for high-performance buildings, and it may be that for some new buildings it has been preferred to the CECB. Finally, the slight over-representation between 2010 and 2017 can be at least partially explained by the advent of a new CECB type (CECBþ) in 2013

Fig. 2. Comparison share of buildings as a function of the construction period between FSO (orange) and the CECB sample (blue).

Fig. 3. Energy label distribution in the CECB as a function of construc­ tion period.

Fig. 1. Heating system type distribution in residential buildings according to the Swiss Federal Statics Office (FSO in orange) and the CECB sample (blue), (Coal ¼ 0.8%).

Fig. 4. a) Energy label distribution by share of buildings and b) by energy reference area (both according to the CECB dataset). 4

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

Table 1 Theoretical and actual final energy consumption in the CECB sample – total and median values by energy label. Energy label

No. Buildings

ERA [km2]

Total Theoretical Consumption [GWh/y]

Total Actual Consumption [GWh/y]

Median Theoretical Consumption [kWh/(m2y)]

Median Actual Consumption [kWh/(m2y)]

A B C D E F G

156 2554 7395 9067 6564 4039 5041

0.10 2.55 7.15 7.64 3.96 1.48 1.28

3.34 123 590 904 646 301 381

3.20 148 671 928 569 231 217

39.4 41.9 78.9 121 164 202 308

37.1 50.2 84.5 116 137 151 174

All

34816

24.2

2950

2770

128

113

mandatory in some cantons to obtain subsides when retrofitting. Also the number of buildings per canton was compared. Some dis­ parities between cantons were found, mostly due to differences in reg­ ulations. However, it is important to highlight that the regulatory situation is rapidly changing, which will affect the future evolution of the dataset. To conclude, despite some differences in representation between cantons, the CECB dataset used for this study can by and large be considered as representative for the Swiss residential building stock. Fig. 3 shows age and energy label distribution of the CECB dataset. Fig. 4 shows the distribution of the number of buildings and the corresponding total energy reference area as function of the energy label in the CECB sample. The distributions are similar for the number of buildings and ERA, with the notable exception of F and G, for which there are more buildings relative to the ERA, indicating that these label categories are primarily found among smaller residences; to a lesser extent this also seems to be the case for labels B and E. The label G in Fig. 4a does not follow the normal distribution ex­ pected from Fig. 4b. This may be caused by the fact that this is the most heterogeneous group defined only by the minimum energy demand (i.e. no maximum demand) and therefore includes a large range of building performance levels.

between theoretical and actual consumption as a function of the energy label. As expected, the theoretical consumption is steadily decreasing with the improvement of the energy performance of the building. The actual consumption behaves quite differently, showing only little dif­ ference in energy consumption for the low ratings (E, F, and G labels) and a substantial reduction similar to the theoretical one for better ratings (A, B, C, and D labels). This once more suggests that an advancement from an energy label G to F or E would not deliver the expected improvement and that it is instead necessary to reach at least a label D to achieve significant actual energy savings. This aspect is particularly relevant when public subsidies for energy retrofit are related to the number of label steps taken. Fig. 6 furthermore shows that buildings with low performance (G, F, and E labels) consume very substantially to moderately less than pre­ dicted. Vice versa, buildings with higher performance (B and C) consume marginally more than predicted. This result supports previous finding in the literature (Delghust et al., 2015; Majcen et al., 2013; �lez, 2013; Raynaud, 2014; Sharpe Merzkirch et al., 2014; Ramallo-Gonza and Shearer, 2013). A-label buildings appear to consume slightly less than predicted, this is further discussed in the following section. Furthermore, it needs to be emphasised that overall the absolute dif­ ference between theoretical and actual energy consumption remains small for all labels A to D. These observations have been tested using a dependent-means t-test (Field, 2009). The difference between actual and theoretical consump­ tion per energy label was identified as statistically significant (p < 0.05; Table 2). Moreover, given the different sample size of each energy label, also the effect size (i.e. an objective and standardized measure of the magnitude of an observed effect) was calculated. As can be seen in Table 2, especially for the E, F, and G labels (r > 0.5) the test yields a fairly large effect size. Therefore, as well as being statistically signifi­ cant, this effect is large and so represents a substantive finding. Fig. 7 shows that a similar pattern as displayed in Fig. 6 is also found across the construction periods: for old buildings actual consumption is lower than theoretical consumption, while buildings built after the year 2000 are consuming somewhat more than expected. It should be noted that the values close to zero in Fig. 7 are always positive even if small, and that manual inspection of a selection of these cases did not reveal any clear data corruption or other markers that the record was invalid. A possible explanation is that some buildings are from the same construction period but have been retrofitted, and therefore presenting a lower energy consumption than other buildings of that period. Nevertheless, the graph confirms that old and less per­ forming buildings consume significantly less energy than expected. These results were found to be statistically meaningful (p < 0.001) through a dependent-means t-test on actual and theoretical consumption per construction period (full results of the t-test in Appendix C).

4. Results and discussion 4.1. Energy consumption difference Total actual final energy consumption for thermal use (2.77 TWh/y) was 6% lower (0.18 TWh/y lower) than the total theoretical consump­ tion (2.95 TWh/y) in the CECB building sample. The detailed results of aggregate and median consumptions as a function of the energy labels are presented in Table 1. The total theoretical and actual consumption are weighted by the shares of the individual label, while the median values are indicative of the typical consumption of a building with that energy label. Fig. 5 il­ lustrates the total actual (in orange) and theoretical (in blue) con­ sumption by energy label (totals according to Table 1). The figure clearly displays the different weights of the energy ratings in energy con­ sumption, with “central” labels C, D, and E dominating total energy consumption (73% of the total). The pattern displayed in Fig. 5 is determined by both the very different ERA distribution (Fig. 4b) and the average energy consumption per label (in kWh/(m2y)). F and G labels jointly account for 23% of total consumption. Build­ ings with these energy labels consume far less energy than expected while the differences are relatively limited for the other labels. Energy retrofitting only F and G labelled buildings would therefore not be suf­ ficient in order to reach ambitious energy saving targets. At the same time, deep energy retrofitting of buildings with an E, D and especially a C label is more difficult to justify, considering that these are consuming less energy per square meter and the payback times of the investments can be expected to be longer than for buildings with an F or G label. The previous analysis was repeated for normalised final energy consumption per square metre of dwelling. Fig. 6 shows the comparison

4.2. Energy performance gap The energy consumption difference in the building sample high­ lighted in section 4.1 is caused by the existence of a significant EPG per building. In this work, the median EPG was found to be 11% (Equation 5

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

Fig. 5. Sum of building energy consumption as a function of energy label (ATheoretical: 3.3; A-Actual:3.2).

Fig. 7. Theoretical and Actual consumption per construction period (tabulated values in Table D2).

(Feist et al., 2003; Peper and Feist, 2015). However, the results con­ cerning the A-label buildings should be treated with caution, as these buildings were poorly represented in the sample: only 156 buildings with A-label were present in the dataset, equal to the 0.5% of the total sample. It is important to note the minimum and maximum values in Fig. 8, which show that the energy consumption of some buildings is actually very different than expected. This spread of values also demonstrates that there is a strong overlap in energy use across the ratings. This means that using only the energy rating as predictor of actual energy con­ sumption is subject to high uncertainty. Using theoretical consumption values as proxy for real performance entails the risk of incorrect assessment, e.g. as the result of discrepancies in operating conditions, malfunctioning of technical systems, and the neglect of a range of loads (e.g. plug loads). Such a rough approach should be used in policy making only to evaluate the entire national building stock (here, the extremes can be expected to balance out). However, it should be kept in mind not to use these theoretical calculations as reference values for actual results (as pointed out by Burman et al., 2014). Table 3 gives the median values of the EPG for each energy label. When interpreting these values and likewise Fig. 8, it should be taken into account that, across the labels, one percent point represents very different energy consumption values in terms of kWh/(m2y). For example, in label E an EPG of 15% corresponds to a difference between actual and theoretical of 27 kWh/(m2y), while in label B, a similar EPG with inverse sign, þ12%, corresponds to an absolute difference of only þ8 kWh/(m2y). These values are smaller than those reported in existing case studies in Switzerland (Khoury et al., 2018; Thaler and Kellen­ berger, 2017), most likely because this is the first stock-level assessment

Fig. 6. Theoretical and Actual consumption per energy label (tabulated values in Table D1)11.

1). This EPG value is not equal to the difference in energy consumption across the sample ( 6%, see section 4.1) due to weighting of the EPG values per building with the respective ERAs and the specific energy consumption per meter square.2 Fig. 8 expresses the difference between theoretical and actual con­ sumption in terms of EPG (Equation 1). The EPG shifts from a large negative value to a relatively moderate positive one as the energy rating improves (with the exception of label A, see below). That is to say, the buildings go from consuming less than expected (negative EPG) to consuming more than expected (positive EPG for label B). It is also important to note that there is a large spread between the minimum and maximum values for every label. The trend for A-label buildings is less clear. Although many buildings showed a positive EPG the median was negative, which could support the theory that the most efficient buildings are more robust to the EPG Table 2 t-test for actual and theoretical consumption per energy label (*p < 0.05, **p < 0.001). Energy label

t-test

Sample size

Effect size [r]

A B C D E F G

2.587* 16.41** 14.97** 15.35** 49.83** 60.19** 79.84**

156 2554 7395 9067 6564 4039 5041

0.202 0.308 0.172 0.159 0.524 0.688 0.747

2 The difference in energy consumption ( 6%) is obtained summing up the absolute theoretical and actual consumption values across the energy labels (Fig. 6), and then calculating their absolute difference. On the other hand, the median EPG ( 11%) is calculated using Equation 1 for each building in the CECB sample, and then the median is calculated within the resulting EPG distribution.

Fig. 8. Energy performance gap per energy label (tabulated values in Table D3). 6

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

Table 3 Energy performance gap per energy label.

Table 4 Energy performance gap per construction period.

Energy label

Median EPG [%]

Construction period

Median EPG [%]

A B C D E F G

6.19 12.5 3.57 5.22 15.4 24.3 40.4

Before 1919 1919–1945 1945–1960 1960–1970 1970–1980 1980–1990 1990–2000 2000–2010 After 2010

17.1 16.2 16.0 12.1 13.1 6.06 3.15 9.71 10.5

(rather than focusing on small numbers of high efficiency buildings). Fig. 9 presents the EPG per construction period, showing a similar pattern as displayed in Fig. 8. The EPG shifts from a negative value to a positive one as the construction period increases. It is also important to note that again there is a large spread between the minimum and maximum values for every construction period, which can be partially justified by a combination of old and renovated buildings. Table 4 gives the median values of the EPG for each construction period, calculated using Equation 1. The direction of the EPG between old and new buildings is similar to that between low and highperformance buildings, even if the maximum values are in absolute terms smaller than in Table 3. However, it is important to note that a given construction period contains both retrofitted and non-retrofitted buildings, i.e. an old building is not necessarily characterized by lower energy performance. As mentioned above, the attention should not be limited exclusively to older buildings with a lower energy rating (labels F and G), which contribute only to a limited extent to actual consumption as they are currently performing better than expected and as they have a relatively small ERA (see Figs. 4 and 5, and Table 1). An effective energy policy should instead include the buildings in the middle rating (labels D and E), which contribute substantially to the total actual consumption thanks to their greater ERA. These findings may partly also explain the so-called rebound effect according to which, after a building retrofit, only a portion of the energy reductions estimated are achieved in practice due to changes in occu­

pant behaviour (Druckman et al., 2011). In existing literature (Galvin, 2013; Haas and Biermayr, 2000), savings are often calculated using a mix of theoretical and actual consumption data, therefore the observed EPG between theoretical and actual consumption (independent of en­ ergy retrofit) could explain part of this difference between expected and achieved savings. However, it once more needs to be noted that the EPG (in % terms) for C-label buildings is very small (Fig. 8), that the somewhat larger EPG for B-label buildings (Fig. 8) translates to a small EPG in specific energy terms (between 6 and 8 kWh/(m2y) for label B and C according to Table 1 and that the EPG for A-label buildings is even negative (Fig. 8). 4.3. CO2 emissions per label category Similar analyses to those carried out on energy were also performed for the CO2 emissions. Table 5 presents the aggregate total theoretical and actual CO2 emissions per energy label, as well as the respective median values. Fig. 10 shows the total actual and theoretical CO2 emissions per energy label, to better highlight the weight of each energy label on the overall CO2 emissions. These closely follow the pattern for total energy consumption shown in Fig. 5. As for the energy consumption, the main contributors to CO2 emissions are the buildings in the C, D and E labels. The contribution of B-labels is very small, partly due to their small total ERA (Fig. 4), while the single A-label building remains invisible on this scale. A further important reason for the very low representation of A and B label buildings is that they use a much smaller or even negligible share of fossil fuels (see Fig. 11). The similarity between Fig. 5 (energy) and Fig. 10 (CO2 emissions) is very interesting. A much larger contribution to CO2 emissions would have been expected, for the same amount of energy consumed, from old buildings with worse energy ratings due to their arguably inefficient and polluting heating systems. This effect is attenuated by the installation of new heating systems in old building3. On the other hand, the homoge­ neity of emissions suggests that the heating systems used in medium rating buildings (labels C, D, E) are not substantially different from those used in the lower rating (labels F, G). This is confirmed by the analyses shown in Fig. 11, where the heating systems are presented as a function of the energy label. As shown in Fig. 11, the mix of technologies and fuels used in labels C to G is not very different and the presence of oil and gas boiler is pre­ dominant in these categories, leading to very similar carbon content of the energy supply for these labels of around 0.23 kg CO2-eq/kWh. In contrast, the absence of oil boilers in buildings with labels A and B brings their carbon content to about 0.12 kg CO2-eq/kWh. The results of CO2 emissions per metre square of ERA (Fig. 12) are determined by both annual energy consumption per m2 as a function of

Fig. 9. Energy performance gap per construction period (tabulated values in Table D4).

1 The boxes in Fig. 6 extend from the lower to upper quartile values of the data, with a line at the median. The whiskers extend from the edges of box to indicate the variability outside the interquartile range (IQR ¼ upper - lower). The position of the whiskers (i.e. minimum and maximum) is set to 1.5 * IQR from the edges of the box. Outliers are not reported (this also applies to all the other box-plots presented in this manuscript).

3 A supplementary analysis of the CECB dataset on the installation periods of the heating systems showed that most old buildings are equipped with new boilers, but still powered by fossil fuels (74% of all the heating system installed after 1990 in buildings constructed before that year are gas- and oil-fuelled).

7

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

Table 5 Theoretical and actual CO2 emissions in the CECB sample – total and median values by energy label. Energy label

No. Buildings

ERA [km2]

Total Theoretical CO2 Emissions [kt CO2-eq/y]

Total Actual CO2 Emissions [kt CO2-eq/y]

Median Theoretical CO2 Emissions [kg CO2-eq/(m2y)]

Median Actual CO2 Emissions [kg CO2-eq/(m2y)]

A B C D E F G

156 2554 7395 9067 6564 4039 5041

0.10 2.55 7.15 7.64 3.96 1.48 1.28

0.42 17.8 131 237 173 78.7 90.5

0.37 22.1 149 243 152 60.1 51.7

5.02 5.95 16.6 30.4 41.8 50.3 68.7

4.87 7.43 18.0 29.3 35.2 37.6 39.4

All

34816

24.2

728

678

30.1

27.9

gap or the limited use of the building, or to incorrect information in the certificate. More generally, a concern with the CECB is related to the influence of human error of the CECB Experts when evaluating buildings and compiling certificates. A study of Bauer and Kuenlin (2013), studied this potential influence, asking eight experts to produce a certificate for the same buildings given the same information and instruments. The main finding was that the experts repeated the same errors in different anal­ ysis, however these led to the assignment of different energy labels (D instead of C) in only one case. Under the assumption that this small sample is representative, the inspectors’ influence can be considered irrelevant for the overall distribution of the energy labels. Another source of uncertainty is the acquisition of the actual energy data. This consumption value is an average result over three years and corrected for heating degree days relative to the standard year which should eliminate the effect of particularly extreme conditions (particu­ larly cold or hot years) on the yearly consumption. However, this is a manual operation and the responsibility of the CECB Expert and there is no way to verify that this was conducted correctly according to the guidelines. To support the quality of the data, it should on the other hand be noted that more than 7000 buildings have been excluded because they were built in the last three years and are therefore lacking the necessary energy consumption readings. This should ensure that a very high percentage was correctly evaluated. The CECB dataset was shown to be largely representative of the building stock. However, it is known that instead of a CECB label some high-performance buildings in Switzerland have the so-called Minergie label (Minergie, 2010), the requirements of which are comparable to an A-label or better. Fig. 4 includes all buildings certified with CECB but omits high-efficiency buildings which only have Minergie labels (which could be considered to belong in the A-label category). However, at the end of 2018, Minergie certified buildings accounted for only 2.3% of the residential building stock (Minergie, 2018), thus having minimal impact on the overall distribution presented in Fig. 4 and efficiency distribution of the residential stock. Given the limited number of A-label buildings in this dataset, the

Fig. 10. CO2 emissions as a function of energy label (A-Theoretical: 0.42; A-Actual:0.37).

energy label and the change in energy mix (increasing shares of gas at the expense of oil from label G to C and growing shares of heat pumps from E to B). Similar to the energy consumption, the actual CO2 emis­ sions in lower performing buildings are found to be smaller than the theoretical ones. The graph also shows that the CO2 emissions are reduced by almost a third when moving from C to B label, a reduction which is notably more pronounced than for energy consumption (Fig. 6) revealing the high efficiency and low CO2 intensity of the heating sys­ tems installed in new or deeply renovated buildings. 4.4. Limitations The large number of buildings in the dataset used in this study is a significant advantage but also presents certain limitations. The opera­ tions of filtering and cleaning the data were already discussed in section 3.2, nevertheless even in the final sample used in the analyses there is a large range of magnitudes of actual energy consumption for buildings with identical energy rating. This phenomenon could be due to the presence of secondary or holiday houses, which are mostly vacant, and therefore have an actual yearly consumption which is very different from the theoretical. Unfortunately, in these cases it has not been possible to determine if the difference was due to a larger performance

Fig. 12. Theoretical and Actual CO2 emissions per energy label (tabulated values in Table D5).

Fig. 11. Heating system distribution in function of energy label. 8

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

results from their analysis should be treated with caution. A different dataset containing more highly efficiency buildings (e.g. a dataset from Minergie) would be required to better understand the EPG in this kind of building. Moreover it may be necessary to enlarge the scope of the performance gap studies to the total final energy consumption of the building instead of only studying consumption for thermal uses, as several studies have confirmed that appliances and lighting are the most demanding end-use applications for many new highly efficient buildings in temperate climates (Christenson et al., 2006; Frank, 2005; Vuarnoz et al., 2018).

currently are not providing any subsidy. Group III accounts for the 43% of the cantons (11 in total), while group I and II represent 19% each (5 cantons per group). Finally, mixed strategies are applied in the remaining 19% of the cases (5 cantons). Similarly to a previous analysis for Switzerland by Siller et al. (2006) the strategies of the three groups have been applied to the current dataset to create scenarios on the technical potential by 2050. These scenarios are developed using as input the median consumption of the energy label categories and the number of buildings in the stock un­ dergoing energy retrofit. Therefore, these simplified scenarios disregard climate change (that could reduce the heating demand in buildings), population growth (with the consequent increase of ERA), and the potentially changing use of the building by the inhabitants (which could partly reduce the EPG). The outputs are the final energy consumption and the energy retrofit rate that is required to reach the objectives by 2050 (calculated by determining the number of buildings to be retro­ fitted and distributing this number over the years remaining to 2050). The results are summarized in Table 6. A business-as-usual case with no incentives (as for Group III) was calculated for the year 2050, assuming an energy retrofit rate of 0.82% per year in line with the current level (Conferenza Cantonale dei Dir­ ettori dell’Energia CDE, 2016). Both the results for actual and theoret­ ical consumption are very far from the objectives of the NEP scenario. This highlights the inadequacy of the business-as-usual case. Scenario 1 follows the policy of Group I and requires all the buildings to be renovated to B-label, resulting in a high renovation rate of 3.11% per year which does not seem to be a realistic solution for meeting NEP objectives. Scenario 2 implements the policy of Group II and assumes all buildings to be improved by three-label steps. Therefore, all buildings except for A, B, and C labels are renovated. These are equal to 71% (DþEþFþG ¼ 24 711 buildings) of the CECB sample, that reported at the entire residential sector level are equal to 1 228 595 buildings. In order to retrofit this amount of buildings in 32 years (2050–2018), a yearly retrofit rate of 2.22% (71% of the building stock/32 years) is found, which is slightly higher than the one proposed in the NEP. In this scenario, the total difference between actual and theoretical final energy consumption is 11%. This explains why the theoretical energy con­ sumption is around 1 TWh below the NEP objective, while the actual energy consumption is around 2 TWh above it. Instead of reducing final energy demand by 46% (from 53.8 TWh/y to 29.2 TWh/y) it is lowered by 42% (from 53.8 TWh/y to 31.1 TWh/y) as a consequence of the EPG. This scenario highlights that the EPG could somewhat affect goal achievement under the ES-2050. The extent of target failure is not very serious because the differences between theoretical and actual values are relatively small for the labels A to D (see Fig. 6). It is worthwhile to mention that in all the 2050 scenarios, the actual consumption is always higher compared to the expected value. This effect is due to the renovation of the older buildings that are currently consuming less than expected to higher ratings which consume more than expected, and therefore saving less energy. This must be taken into account in order to ensure achievability of the ES-2050 objectives. These scenarios were calculated based on a conservative approach by basing the scenario analysis on median values, and therefore do not consider uncertainty in the results. As can be seen in Fig. 8, buildings can have a large range of EPG within each label, and this could greatly affect the energy savings when transitioning between labels (i.e. if a building transitions from/to values near the median compared to transitioning between extremes of the distribution). Developing scenarios considering all these elements is a topic of ongoing work.

5. Scenarios and incentives The correlation between the negative EPG of low performing build­ ing is noteworthy because the expected energy savings from their renovation may be much less than expected. In light of this issue, the theoretical and actual consumption values are now used to determine how the EPG affects the achievement of the ES-2050 objectives under simplified technical potential scenarios. This model is based on the existing building stock while not representing the dwellings that will be built in the coming years (where there will anyway be no energy ret­ rofitting for the time being). In addition, this section aims to address Article 2a.1.d of the new EPBD (The European Parliament and the Council of the EU, 2018), which calls for an overview of policies and actions to target the worst performing segments of the national building stock. As first step of our scenario analysis, the results obtained from the CECB have been scaled to the entire Swiss residential sector to obtain an estimation of the final energy for thermal use. When scaling the CECB sample (24 million m2) to total residential ERA (470 million m2), a theoretical consumption of 57.4 TWh/y and an actual consumption of 53.8 TWh/y was found. The calculation for Switzerland was realized assuming the same energy label distribution of the CECB sample within the residential sector (considering the representativeness of the CECB dataset, Section 3.3). Therefore, the results were scaled using the me­ dian values of each label for the corresponding m2 of building stock. For example, if B-label ERA corresponds to 10.5% of the CECB sample, and by assumption also to 10.5% of the ERA of the entire residential sector, then 49.4 million m2 (10.5% of the total ERA reported by the FSO) was multiplied by 50 kWh/m2 (median actual consumption of B-label, Table 1) and so on for all other energy labels to obtain the total theo­ retical and actual consumption for the residential sector. These results confirm the representativeness of the CECB dataset, as Swiss authorities have reported the yearly final energy use for space heating and domestic hot water in the residential sector to be 54.7 TWh (Prognos, 2017). While our scaled-up results do not include new buildings with high to very high energy performance (with labels belonging to the Minergie family) these hardly influence total energy demand due to their low share (see section 4.4). The declared target of the New Energy Policy (NEP) for final energy consumption for thermal use in residential buildings by the year 2050 is 29.2 TWh (Prognos, 2012). To reach this objective, a renovation rate of 1.9% per year of the building stock was proposed (Prognos, 2012). Concurrently, each canton in Switzerland is applying various policies to achieve these targets, thereby aiming to minimize the costs (Amstalden et al., 2007). Based on a review of the cantons’ energy strategies, it was found that their approaches to subsidising CECB energy label improvements fall into three main groups. Group I contains all cantons that are providing subsidies only if at least the B-label is achieved. Group II contains all the cantons that provide uniform subsidies for every extra label level gained regardless of the end point. Group III contains all the cantons that

9

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

Table 6 Technical potential scenario for the Swiss residential sector by 2050 (thermal performance only). The energy consumption difference is defined as the difference in actual consumption as a proportion of the theoretical consumption. Scenario

Cantons

Final Energy [TWh/y] Actual

2017 2050

CECB Model Prognos Business as-usual 1 2 NEP

All All Zurich - Uri Basel - Vaud Bern - Geneva All

53.8 47.3 18.8 31.1

Energy consumption difference [%]

Renovation rate [%]

6 0 2 4 11 n/a

n/a n/a 0.82 3.11 2.22 1.9

Theoretical 57.4

54.7

46.2 17.9 28.1

29.2

6. Conclusions and policy implications

between Swiss cantons creates an opportunity for research. Three groups of energy policies linking incentives to energy rating improvement were found among the cantons in Switzerland. These three polices were turned into technical potential scenarios, to assess the feasibility of the ES-2050 targets. The results show that with the business-as-usual approach maintaining current renovation rates for the oldest build­ ings, it is impossible to achieve the objectives set. On the other hand, a more aggressive policy aiming at renovating all buildings to the highest standards (Scenario 1), translates into a renovation rate that seems un­ realistically high (3.1% per year). An intermediate approach involving the improvement by three label steps and achievement of label D as minimum requirement (Scenario 2) is the most promising approach (with a renovation rate of 2.2% per year). Nevertheless the effect of the performance gap has to be taken into account, resulting in a somewhat higher actual consumption than the target according to theoretical en­ ergy demand (31.1 TWh/y instead of 29.2 TWh/y, i.e. by 7% higher). This initial scenario analysis has demonstrated significantly different impacts of cantonal energy policies under the assumptions that they are applied uniformly to the building stock. Moreover, subsidies can be a good option to increase the number of retrofits and to achieve the theoretical energy targets. However, in view of the highlighted EPG, it is suggested to policy makers to incentivise the owner to take more expensive but more energy-efficient measures, giving subsidies only when the improvement of energy performance is equivalent to at least three label steps and when goal achievement has been proven. However, further analysis of these policies options, considering decarbonisation of the heat supply, is needed to determine an optimal energy retrofit strategy which take the EPG into account. The type of permitted and subsidized retrofits is another aspect on which national policies should focus more. Typically, authorities make the compliance with minimum energy performance levels mandatory only if a building, a building unit or a building element is subject to major renovation (Republique et Canton de Geneve, 2017; The European Parliament and the Council of the EU, 2018). For example, in the context of its decar­ bonisation policy, the canton of Basel City has prohibited the replace­ ment oil and gas-fired boilers since January 2018 (Amt für Umwelt und Energie AUE des Kantons Basel-Stadt, 2018). Some Swiss cantons set upper thresholds for energy demand per m2 (e.g. 600 MJ/(m2y) in Geneva; Republique et Canton de Geneve, 2019) beyond which owners are contacted and requested to make modifications. Generally speaking, more coercive policy instruments are being implemented in some loca­ tions while the balance between energy and climate policy goals on the one hand and property protection and grandfathering rules on the other represent a delicate balance. So far, these policies focus on taking measures aiming at lower energy use and CO2 emissions while typically no proof of performance after commissioning is mandatory. In other words, existing policy does not (yet) acknowledge the existence of the EPG. Future policies, for example making use of more novel instruments like Building Renovation Passports (a document outlining a long-term step-by-step renovation roadmap for a specific building; Sesana and Salvalai, 2018), will need to actively tackle this challenge. To conclude, this research suggests that while policy makers should encourage energy retrofitting to reach the ambitious targets of the national energy

This paper has investigated the thermal Energy Performance Gap, defined as difference between a building’s theoretical and actual energy consumption for thermal purposes (space heating and domestic hot water), in the residential sector. The work was based on the official database of the Cantonal Energy Certificate for Buildings that contains data of almost 50 000 buildings, making this study the most compre­ hensive so far conducted in Switzerland and one of the largest in Europe. The CECB reports the energy efficiency of the building, distinguished into classes A to G by means of an energy label, the actual consumption by the energy bills, and the building’s energy requirements according to Swiss standards. Using this information the EPG was quantified for final energy use and CO2 emissions related to the supply of space heating and domestic hot water. The analyses revealed the existence of an EPG for residential build­ ings, with a median of 11% per building. This implies that the actual consumption of final energy is 6% lower than predicted for the entire residential stock. This is due to the large share of buildings with lower energy rating (label E-G) which were found to perform significantly better than predicted. A strong correlation was found between energy label and EPG. For low performing G-labelled buildings, a negative EPG of 40% was determined, while for the higher performing buildings with B-label a positive EPG of þ12% was determined. These results imply that higher performance buildings are consuming somewhat more than predicted while lower performance buildings are consuming significantly less than predicted. Nevertheless, the limited number of Alabel buildings in the CECB dataset and the absence of very high per­ formance buildings (e.g. Minergie) calls for further investigations. A similar pattern to the EPG as a function of the energy labels was found for CO2 emissions. An important finding is that the actual emis­ sions in terms of kg CO2-eq/kWh are very similar between the labels C and G, showing a large drop only for the labels A and B which is not matched by an equally large drop in energy consumption. This high­ lights the contribution of the high efficiency and low emissions of the heating systems installed in these new or deeply renovated buildings, and further indicates a need for energy policy to incentivise reducing carbon emissions in addition to requiring a minimum level of energy efficiency. This finding shows that one of the major limitations of the current certification scheme is the lack of monitoring of the consump­ tion after the certificate has been issued. In this way it becomes very difficult to assess the real performance of a building or the outcome of retrofit and to identify, and possibly close, the EPG. The proposition of monitoring as a technical solution to close the gap has already been discussed by several authors (Guerra-Santin and Tweed, 2015; Menezes et al., 2012; Niu et al., 2016; Sagerschnig, 2015). This implies that a priority for policy makers should be to implement monitoring as a requirement for certification, or mandatory by law for those who are responsible for the work. Instead of implementing monitoring for each and every building, sample monitoring may be sufficient. Finally, in view of the existence of the EPG, this study provides a first analysis of the feasibility of achieving the objectives of the national Energy Strategy 2050. In this regard, the diversity of policy approaches 10

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

strategy, a good strategy should aim not only to bring buildings from the lowest to the highest rating, but also to improve the mid-range by setting minimum intermediate targets (D-label). At the same time future research should be conducted on the EPG in high performing buildings, in order to arrive at a more complete understanding of the consequences of deep renovations of the current building stock.

Acknowledgements This research was conducted under the GAPxPLORE project (“Energy performance gap in existing, new and renovated buildings – Learning from large-scale datasets”) funded by the Swiss Federal Office of Energy (SFOE; contract number SI/501518-01) as well as under the Swiss Competence Center for Energy Research on Future Energy Efficient Buildings & Districts (SCCER FEEB&D) of the Swiss Innovation Agency Innosuisse. We gratefully thank Kai Streicher, Achim Geissler, and Karine Wesselmann for their valuable contributions.

Declaration of competing interest There are no known conflicts of interest.

Appendix A The standard condition of use of the building for the theoretical calculation are given in the Swiss norms SIA 380/1 and SIA 2024, based on the EU norm EN ISO 13790, and summarized in Table A1. Table A.1 Standard conditions of use. Operational parameters

Unit

Standard conditions

Indoor temperature Surface per person Thermal gain per person Metabolic activity Days of use per year Occupied hours Appliance thermal gain Appliance use hours Lighting thermal gain Lighting use hours Air change rate Hot water demand per person

�

C m2/p W/p met d/y h/d W/m2 h/d W/m2 h/d m3/(m2*h) l/(d*p)

20 40 70 1.2 365 12 8 6.1 2.7 7 0.7 35

Appendix B The uncertainty in the actual energy consumption is the sum of the uncertainties in taking the average of the measurements over three consecutive years and in the energy carriers’ readings. For the first part we consider this uncertainty balanced out by the climate correction of the theoretical calculation. For the second part, this research assumes that the error in the electricity meter reading is negligible. However, the uncertainty in gas, oil, and wood consumption could be significant because they are measured through volumetric flow and converted to an energy value. Several sources have been reviewed on this topic (Chambers, 2017; IEA and Eurostat, 2004; Lander, 2012; Office of Gas and Electricity Markets, 2000), and concluded that for this kind of analysis the uncertainty in the conversion factors has a negligible effect on the final actual energy consumptions. Therefore, the standard values proposed by the Swiss norms are used and reported in Table B1. Table B.1 Density and calorific power of the different energy carriers considered in this study, from the Swiss norm SIA 2031. Energy carrier

Density [kg/m3]

Calorific power [MJ/kg]

Oil Coal Wood, pieces Wood, chips Wood, pellets Natural Gas Biogas

840 700 350 170 660 0.76 1.3

44.8 21.2 19.9 19.9 20.2 40.3 23

Appendix C Dependent-means t-test on actual and theoretical consumption per construction period. The results are statistically meaningful. Moreover, given the different size of the sample of each construction period, also the effect size was calculated. As can be seen in Table C1, especially for buildings build

11

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

before the 1980, the test yields a fairly large effect size. Therefore, as well as being statistically significant, this effect is large and so represents a substantive finding. Table C.1 t-test for actual and theoretical consumption per construction period (*p < 0.001). Construction period

t-test

Sample size

Effect size [r]

Before 1919 1919–1945 1945–1960 1960–1970 1970–1980 1980–1990 1990–2000 2000–2010 After 2010

41.25* 32.44* 38.52* 32.47* 34.97* 17.74* 6.217* 12.26* 8.008*

5355 3142 4613 4724 5845 5513 2705 2308 611

0.491 0.501 0.493 0.427 0.416 0.232 0.119 0.247 0.308

Appendix D Tabulated values for the box-plots presented in this manuscript (Figs. 6–9 and 12). Table D.1 Theoretical and Actual consumption [kWh/(m2y)] per energy label. Energy label

No. Buildings

Consumption

Min.

25%

50%

75%

Max.

A

156

B

2554

C

7395

D

9067

E

6564

F

4039

G

5041

Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual

2.57 3.70 7.80 2.92 14.3 9.60 28.9 22.1 49.4 26.8 69.5 20.9 82.2 8.07

18.3 15.7 25.2 29.8 52.3 56.5 94.4 90.8 131 106 152 107 209 110

39.4 37.1 41.9 50.2 78.9 84.5 121 116 164 137 202 151 308 174

52.3 49.8 65.2 74.8 101 106 145 138 194 165 244 186 372 216

87.6 86.4 117 140 181 190 228 214 303 263 395 312 671 415

Table D.2 Theoretical and Actual consumption [kWh/(m2y)] per construction period. Construction period

No. Buildings

Consumption

Min.

25%

50%

75%

Max.

Before 1919

5355

1919–1945

3142

1945–1960

4613

1960–1970

4724

1970–1980

5845

1980–1990

5513

1990–2000

2705

2000–2010

2308

After 2010

611

Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual

8.60 6.04 2.57 5.88 8.40 2.92 8.90 3.70 8.50 5.85 7.80 8.93 10.4 4.05 3.20 4.71 6.30 3.07

108 85.6 119 99.4 115 100 112 105 102 90.0 77.3 69.2 67.3 64.1 35.2 40.4 21.1 23.5

154 121 168 133 169 135 162 137 142 122 103 96.2 90.9 87.6 61.6 69.8 38.3 42.5

219 158 231 173 238 173 223 171 192 154 131 121 114 112 83.4 90.9 55.2 66.1

671 415 632 411 670 414 651 411 661 408 537 312 508 353 275 200 106 166

Table D.3 Energy performance gap [%] per energy label. Energy label

No. Buildings

A B C D E F G

156 2554 7395 9067 6564 4039 5041

Min.

25%

82.3 91.7 90.9 83.7 85.3 90.1 98.3

24.4 7.35 13.1 20.1 30.6 40.4 59.0

12

50%

75%

Max.

6.19 12.5 3.57 5.22 15.4 24.3 40.4

11.4 43.9 28.9 13.1 0.34 8.34 19.6

697 1210 688 333 213 119 205

S. Cozza et al.

Energy Policy xxx (xxxx) xxx

Table D.4 Energy performance gap [%] per construction period. Construction period

No. Buildings

Before 1919 1919–1945 1945–1960 1960–1970 1970–1980 1980–1990 1990–2000 2000–2010 After 2010

5355 3142 4613 4724 5845 5513 2705 2308 611

Min.

25%

97.6 97.4 98.3 95.2 96.3 96.5 91.7 89.6 88.8

37.7 34.3 34.8 30.8 31.8 24.4 20.2 10.6 7.07

50%

75%

Max.

17.1 16.1 16.0 12.1 13.1 6.06 3.15 9.71 10.5

0.73 0.87 1.27 6.89 5.91 14.2 17.3 42.7 33.4

697 430 559 1210 426 688 569 652 468

Table D.5 Theoretical and Actual CO2 emissions [kg CO2-eq/(m2y)] per energy label. Energy label

No. Buildings

Consumption

Min.

25%

50%

75%

Max.

A

156

B

2554

C

7395

D

9067

E

6564

F

4039

G

5041

Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual Theoretical Actual

0.26 0.22 0.27 0.00 0.27 0.39 0.65 0.55 1.24 0.70 1.69 0.80 2.26 0.27

1.11 1.34 1.77 2.03 5.31 6.07 23.4 22.6 33.9 26.6 30.7 18.1 20.2 12.1

5.02 4.87 5.95 7.43 16.6 18.0 30.4 29.3 41.8 35.2 50.3 37.6 68.7 39.4

8.75 7.89 9.85 11.2 24.0 25.8 39.4 37.3 55.1 46.4 70.2 52.3 97.1 58.1

25.5 24.0 35.1 40.4 54.2 57.0 68.1 64.5 89.9 79.3 107 93.9 202 125

References

Sheldrick, B., Csoknyai, T., Hrabovszky-Horv� at, S., Szendr} o, G., Soto, L., Leticia, F., Madrigal, O., Serrano, B., Serghides, D., Dimitriou, S., Katafygiotou, M., Nieboer, N., Filippidou, F., Visscher, H., Brattebø, H., Sandberg, N.H., Vestrum, M.I., Sartori, I., Popovi�c, M.J., 2016. Monitor Progress towards Climate Targets in European Housing Stocks. Final Project Report EPISCOPE. de Wilde, P., 2014. The gap between predicted and measured energy performance of buildings: a framework for investigation. Autom. Construct. 41, 40–49. https://doi. org/10.1016/j.autcon.2014.02.009. Delghust, M., Roelens, W., Tanghe, T., De Weerdt, Y., Janssens, A., 2015. Regulatory energy calculations versus real energy use in high-performance houses. Build. Res. Inf. 43, 675–690. https://doi.org/10.1080/09613218.2015.1033874. Druckman, A., Chitnis, M., Sorrell, S., Jackson, T., 2011. Missing carbon reductions? Exploring rebound and backfire effects in UK households. Energy Policy 49, 778. https://doi.org/10.1016/j.enpol.2012.06.045. EnDK, 2016. Facteurs de pond�eration nationaux pour l’�evaluation des batiments. Bundesamt für Energie (BFE). EU Parliament, 2014. Energy Efficiency and its contribution to energy security and the 2030 framework for climate and energy policy, pp. 1–8. SWD(2013)93. Feist, W., Peper, S., Kah, O., von Oesen, M., 2003. Climate Neutral Passive House Estate in Hannover-Kronsberg: Construction and Measurement Results. PEP Project Information. Field, A., 2009. Discovering Statiscs Using SPSS. SAGE. Frank, T., 2005. Climate change impacts on building heating and cooling energy demand in Switzerland. Energy Build. 37, 1175–1185. https://doi.org/10.1016/j. enbuild.2005.06.019. Frei, B., Sagerschnig, C., Gyalistras, D., 2018. ParkGap – Performance Gap Geb€ aude. FSO, 2017. In: Construction and Housing - Key Figures. Federal Population Census, Buildings and Dwellings Statistics. Gaetani, I., Hoes, P.J., Hensen, J.L.M., 2016. Occupant behavior in building energy simulation: towards a fit-for-purpose modeling strategy. Energy Build. 121, 188–204. https://doi.org/10.1016/j.enbuild.2016.03.038. Galvin, R., 2013. Making the ‘rebound effect’ more useful for performance evaluation of thermal retrofits of existing homes: defining the ‘energy savings deficit’ and the ‘energy performance gap. Energy Build. 69, 515–524. https://doi.org/10.1016/j. enbuild.2013.11.004. Grossmann, D., Galvin, R., Weiss, J., Madlener, R., Hirschl, B., 2016. A methodology for estimating rebound effects in non-residential public service buildings: case study of four buildings in Germany. Energy Build. 111, 455–467. https://doi.org/10.1016/j. enbuild.2015.11.063. Guerra-Santin, O., Tweed, C.A., 2015. In-use monitoring of buildings: an overview of data collection methods. Energy Build. 93, 189–207. https://doi.org/10.1016/j. enbuild.2015.02.042. Haas, R., Biermayr, P., 2000. The rebound effect for space heating empirical evidence from Austria. Energy Policy 28, 403–410. https://doi.org/10.1016/S0301-4215(00) 00023-9.

Ahern, C., Norton, B., 2019. Thermal energy refurbishment status of the Irish housing stock. Energy Build. 202, 109348. https://doi.org/10.1016/j.enbuild.2019.109348. Amstalden, R.W., Kost, M., Nathani, C., Imboden, D.M., 2007. Economic potential of energy-efficient retrofitting in the Swiss residential building sector: the effects of policy instruments and energy price expectations. Energy Policy 35, 1819–1829. https://doi.org/10.1016/j.enpol.2006.05.018. Amt für Umwelt und Energie AUE des Kantons Basel-Stadt, 2018. F€ orderbeitr€ age für Energiesparmassnahmen im Kanton Basel-Stadt 12 S. Andaloro, A.P.F., Salomone, R., Ioppolo, G., Andaloro, L., 2010. Energy certification of buildings: a comparative analysis of progress towards implementation in European countries. Energy Policy 38, 5840–5866. https://doi.org/10.1016/j. enpol.2010.05.039. Andri�c, I., Pina, A., Ferr~ ao, P., Fournier, J., Lacarri� ere, B., Le Corre, O., 2017. HIT2GAP: towards a better building energy management. Energy Procedia 122, 895–900. https://doi.org/10.1016/j.egypro.2017.07.399. Arcipowska, A., Rapf, O., Faber, M., Fabbri, M., Tigchelaar, C., Boermans, T., SurmeliAnac, N., Pollier, K., Dal, F., Sebi, C., Kar� asek, J., 2016. Support for Setting up an Observatory of the Building Stock and Related Policies. Buildings Performance Institute Europe (BPIE). Bauer, M., Kuenlin, A., 2013. Bewertung des Experteneinflusses auf die GEAK®Klassifikation eines Geb€ audes. Direktion für Energie (DiREN). Burman, E., Mumovic, D., Kimpian, J., 2014. Towards measurement and verification of energy performance under the framework of the European directive for energy performance of buildings. Energy 77, 153–163. https://doi.org/10.1016/j. energy.2014.05.102. Calì, D., Osterhage, T., Streblow, R., Müller, D., 2016. Energy performance gap in refurbished German dwellings: lesson learned from a field test. Energy Build. 127, 1146–1158. https://doi.org/10.1016/j.enbuild.2016.05.020. Cayre, E., Allibe, B., Laurent, M.-H., Osso, D., 2011. There are people in the house! How the results of purely technical analysis of residential energy consumption are misleading for energy policies. In: ECEEE 2011 Summer Study - Energy Efficiency First. The Foundation of a Low-Carbon Society. Chambers, J., 2017. Developing a rapid, scalable method of thermal characterisation for UK dwellings using smart meter data. PhD. Thesis. UCL Energy Institute. Christenson, M., Manz, H., Gyalistras, D., 2006. Climate warming impact on degree-days and building energy demand in Switzerland. Energy Convers. Manag. 47, 671–686. https://doi.org/10.1016/j.enconman.2005.06.009. Conferenza Cantonale dei Direttori dell’Energia CDE, 2016. Il Programma Edifici nel 2016 - Rapporto annuale. Conferenza Cantonale dei Direttori dell’Energia CDE, 2014. Modello di prescrizioni energetiche dei cantoni (MoPEC). Dascalaki, E., Balaras, C., Droutsa, P., Kontoyiannidis, S., Vandevelde, B., Cuypers, D., Vimmr, T., Villatoro, O., Prague, S.-K., Republic, C., Hanratty, M., Badurek, M.,

13

S. Cozza et al.

Energy Policy xxx (xxxx) xxx Ramallo-Gonz� alez, A.P., 2013. Modelling, Simulation and Optimisation Methods for Low-Energy Buildings. PhD. Thesis - University of Exeter. Raynaud, M., 2014. Evaluation ex-post de l’efficacit�e de solutions de r� enovation �energ� etique en r� esidentiel. PhD. Thesis - ParisTech. Paris Institute of Technology. Reimann, W., Buhlmann, E., Lehmann, M., 2016. Erfolgskontrolle Geb€ audeenergiestandards 2014-2015. Bundesamt für Energie (BFE). Republique et Canton de Geneve, 2019. Optmiser la consommation de chaleur d’un batiment. Republique et Canton de Geneve, 2017. Crit� eres de subvention 2017 pour le canton de Gen� eve. Risholt, B., Berker, T., 2013. Success for energy efficient renovation of dwellings—Learning from private homeowners. Energy Policy 61, 1022–1030. https://doi.org/10.1016/j.enpol.2013.06.011. Sagerschnig, C., 2015. Was ist der «Performance Gap»? IBPSA-CH Working Group. Schr€ oder, F.P., Papert, O., Boegelein, T., Navarro, H., Mundry, B., 2014. Reale Trends des spezifischen Energieverbrauchs und repr€ asentativer Wohnraumtemperierung bei steigendem Modernisierungsgrad im Wohnungsbestand. In: Bauphysik, pp. 309–324. Schuler, A., Weber, C., Fahl, U., 2000. Energy consumption for space heating of WestGerman households: empirical evidence, scenario projections and policy implications. Energy Policy 28, 877–894. https://doi.org/10.1016/S0301-4215(00) 00074-4. Sesana, M.M., Salvalai, G., 2018. In: A Review on Building Renovation Passport: Potentialities and Barriers on Current Initiatives. Energy Build. Sharpe, T., Shearer, D., 2013. Adapting the Scottish tenement to 21st century standards: an evaluation of the performance enhancement of a 19th century “Category B” listed tenement block in Edinburgh. J. Cult. Herit. Manag. Sustain. Dev. 3, 55–67. https:// doi.org/10.1108/20441261311317400. SIA, 2016. SIA 2031 - Energy Certificate for Buildings. Swiss Society of Engineers and Architects (SIA). SIA, 2015. SIA 380 - Basi per il calcolo energetico di edifici. Swiss Society of Engineers and Architects (SIA). Siller, T., Kost, M., Imboden, D., 2006. Long-term energy savings and greenhouse gas emission reductions in the Swiss residential sector. Energy Policy 35, 529–539. https://doi.org/10.1016/j.enpol.2005.12.021. Sunikka-Blank, M., Galvin, R., 2012. Introducing the prebound effect: the gap between performance and actual energy consumption. Build. Res. Inf. 40, 260–273. https:// doi.org/10.1080/09613218.2012.690952 To. Sutherland, G., Audi, P.G., Lacourt, A., Deliyannis, A., Sotiropoulos, D., Poseidon, K., Mcelmuray, P., Skrivanou, M., Koutsou, S., Davis, M.F., Fytrou, A., Tsagkla, M., 2015. Implementing the energy performance of building directive (EPBD). Intelligent Energy Europe Programme. Swiss Federal Office of Energy, 2018. Energy strategy 2050 once the new energy act is in force. Bundesamt für Energie (BFE). Thaler, L., Kellenberger, D., 2017. Addressing gaps: user on behavior and sufficiency in the planning and operation phase of a 2000-Watt Site. Energy Procedia 122, 961–966. https://doi.org/10.1016/j.egypro.2017.07.440. The European Parliament and the Council of the EU, 2018. Directive 2018/844/EU Energy performance of buildings. Off. J. Eur. Union 2018, 75–91. The European Parliament and the Council of the EU, 2003. Directive 2002/91/EC of the European parliament of the council of 16 December 2002 on the energy performance of buildings. In: Official Journal of the European Communities, vol. 2003, pp. L1/ 65–71. Tian, W., Heo, Y., de Wilde, P., Li, Z., Yan, D., Park, C.S., Feng, X., Augenbroe, G., 2018. A review of uncertainty analysis in building energy assessment. Renew. Sustain. Energy Rev. 93, 285–301. https://doi.org/10.1016/j.rser.2018.05.029. Van den Brom, P., Meijer, A., Visscher, H., 2017. Performance gaps in energy consumption: household groups and building characteristics. Build. Res. Inf. 46, 54–70. https://doi.org/10.1080/09613218.2017.1312897. Van Dronkelaar, C., Dowson, M., Spataru, C., Mumovic, D., 2016. A review of the energy performance gap and its underlying causes in non-domestic buildings. Front. Mech. Eng. 1, 1–14. https://doi.org/10.3389/fmech.2015.00017. Vuarnoz, D., Cozza, S., Jusselme, T., Magnin, G., Schafer, T., Couty, P., Niederhauser, E.L., 2018. Integrating hourly life-cycle energy and carbon emissions of energy supply in buildings. Sustain. Cities Soc. 43, 305–316. https://doi.org/10.1016/j. scs.2018.08.026. World Bank, 2018. GDP Per Capita PPP - Current International $. World Bank national accounts data, and OECD National Accounts data files. Wyss, S., H€ assig, W., 2016. UFELD: Feldmessungen von U-Werten zur Überprüfung der im Geb€ audeenergieausweis (GEAK) hinterlegten U-Werte. Bundesamt für Energie (BFE). Zgraggen, J.-M., 2010. B^ atiments r� esidentiels locatifs � a haute performance � energ� etique : objectifs et r� ealit� es. PhD. Thesis. University of Geneva. Zhang, Y., Bai, X., Mills, F.P., Pezzey, J.C.V., 2018. Rethinking the role of occupant behavior in building energy performance: a review. Energy Build. 172, 279–294. https://doi.org/10.1016/j.enbuild.2018.05.017. Zou, P.X.W., Xu, X., Sanjayan, J., Wang, J., 2018. Review of 10 years research on building energy performance gap : life-cycle and stakeholder perspectives. Energy Build. 178, 165–181. https://doi.org/10.1016/j.enbuild.2018.08.040.

Hens, H., Parijs, W., Deurinck, M., 2010. Energy consumption for heating and rebound effects. Energy Build. 42, 105–110. https://doi.org/10.1016/j.enbuild.2009.07.017. Hoffmann, C., Geissler, A., 2017. The prebound-effect in detail: real indoor temperatures in basements and measured versus calculated U-values. Energy Procedia 122, 32–37. https://doi.org/10.1016/J.EGYPRO.2017.07.301. Hughes, M., Palmer, J., Cheng, V., Shipworth, D., 2015. Global sensitivity analysis of England’s housing energy model. J. Build. Perform. Simul. 8, 283–294. https://doi. org/10.1080/19401493.2014.925505. IEA, Eurostat, 2004. Energy Statistics Manual. International Energy Agency (IEA). Iglewicz, B., Hoaglin, D., 1993. Volume 16: how to detect and handle outliers. In: The ASQC Basic References in Quality Control: Statistical Techniques. KBOB, 2016. Donn�ees des �ecobilans dans la construction 2009/1:2016. €ffentlichen Koordinationskonferenz der Bau- und Liegenschaftsorgane der o Bauherren. Kelly, S., 2011. Do homes that are more energy efficient consume less energy?: a structural equation model of the English residential sector. Energy 36, 5610–5620. https://doi.org/10.1016/j.energy.2011.07.009. Khoury, J., Alameddine, Z., Hollmuller, P., 2017. Understanding and bridging the energy performance gap in building retrofit. Energy Procedia 122, 217–222. https://doi. org/10.1016/j.egypro.2017.07.348. Khoury, J., Hollmuller, P., Lachal, B., 2016. Energy performance gap in building retrofit: characterization and effect on the energy saving potential. In: 19. Status-Seminar «Forschen Für Den Bau Im Kontext von Energie Und Umwelt». http://archive-ouve rte.unige.ch/unige:86086. Khoury, J., Hollmuller, P., Lachal, B., Schneider, S., Lehmann, U., 2018. COMPARE RENOVE: du catalogue de solutions � a la performance r�eelle des r� enovations � energ� etiques. Office f�ed� eral de l’� energie (OFEN). Kragh, J., Rose, J., Knudsen, H.N., Jensen, O.M., 2017. Possible explanations for the gap between calculated and measured energy consumption of new houses. Energy Procedia 132, 69–74. https://doi.org/10.1016/j.egypro.2017.09.638. La Fleur, L., Moshfegh, B., Rohdin, P., 2017. Measured and predicted energy use and indoor climate before and after a major renovation of an apartment building in Sweden. Energy Build. 146, 98–110. https://doi.org/10.1016/j. enbuild.2017.04.042. Lander, D.F., 2012. Accuracy of CV Determination Systems for Calculation of FWACV. davelanderconsulting. Lehmann, U., Khoury, J., Patel, M.K., 2017. Actual energy performance of student housing: case study, benchmarking and performance gap analysis. Energy Procedia 122, 163–168. https://doi.org/10.1016/J.EGYPRO.2017.07.339. Loucari, C., Taylor, J., Raslan, R., Oikonomou, E., Mavrogianni, A., 2016. Retrofit solutions for solid wall dwellings in England: the impact of uncertainty upon the energy performance gap. Build. Serv. Eng. Technol. 37, 614–634. https://doi.org/ 10.1177/0143624416647758. Majcen, D., 2016. Predicting Energy Consumption and Savings in the Housing Stock: A Performance Gap Analysis in the Netherlands. PhD. Thesis. Delft University of Technology, Faculty of Architecture and the Built Environment. Majcen, D., Itard, L., Visscher, H., 2013. Theoretical vs. actual energy consumption of labelled dwellings in The Netherlands: discrepancies and policy implications. Energy Policy, Decades of {Diesel} 54, 125–136. https://doi.org/10.1016/j. enpol.2012.11.008. Meijer, A., 2017. Results of the TRIME project. Intelligent Energy Europe Programme. Menezes, A.C., Cripps, A., Bouchlaghem, D., Buswell, R., 2012. Predicted vs. actual energy performance of non-domestic buildings: using post-occupancy evaluation data to reduce the performance gap. Appl. Energy 97, 355–364. https://doi.org/ 10.1016/j.apenergy.2011.11.075. Merzkirch, A., Hoos, T., Maas, S., Scholzen, F., Waldmann, D., 2014. Wie genau sind unsere Energiep€ asse?. In: Bauphysik, pp. 40–43. Minergie, 2018. Standard Minergie. Minergie, 2010. The MINERGIE Standard for Buildings. Niu, S., Pan, W., Zhao, Y., 2016. A virtual reality integrated design approach to improving occupancy information integrity for closing the building energy performance gap. Sustain. Cities Soc. 27, 275–286. https://doi.org/10.1016/j. scs.2016.03.010. Office of Gas and Electricity Markets, 2000. Gas Energy Measurement: A Consultation Document. Measuring Instruments Directive. Peper, S., Feist, W., 2015. Die Energieeffizienz des Passivhaus-Standards: Messungen best€ atigen die Erwartungen in der Praxis. Passivhaus-Institut. Petersen, S., Hviid, C.A., 2012. The European energy performance of buildings directive: comparison of calculated and actual energy use in a Danish office building. In: IBPSA-England; Building Simulation and Optimization Conference, pp. 43–48. Pietropaoli, B., Pusceddu, D., Delaney, K., Pesch, D., 2014. TRIBUTE, Take the energy bill back to the promised building performance. Deliverable: D2.vol. 2, Seventh Framework Programme. Prognos, 2017. Der Energieverbrauch der Privaten Haushalte 2000 - 2016. Bundesamt für Energie (BFE), 31-27264. Prognos, 2012. Die Energieperspektiven für die Schweiz bis 2050. Bundesamt für Energie (BFE).

14