Building Research & Information

ISSN: 0961-3218 (Print) 1466-4321 (Online) Journal homepage: http://www.tandfonline.com/loi/rbri20

Performance gaps in energy consumption:

household groups and building characteristics

Paula van den Brom, Arjen Meijer & Henk Visscher

To cite this article: Paula van den Brom, Arjen Meijer & Henk Visscher (2017): Performance
gaps in energy consumption: household groups and building characteristics, Building Research &
Information, DOI: 10.1080/09613218.2017.1312897

To link to this article: http://dx.doi.org/10.1080/09613218.2017.1312897

© 2017 The Author(s). Published by Informa

UK Limited, trading as Taylor & Francis
Group

Published online: 04 May 2017.

Submit your article to this journal

Article views: 125

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=rbri20

Download by: [80.82.77.83] Date: 09 May 2017, At: 05:48

 BUILDING RESEARCH & INFORMATION
2017
https://doi.org/10.1080/09613218.2017.1312897

RESEARCH PAPER

Performance gaps in energy consumption: household groups and building

characteristics

Paula van den Brom, Arjen Meijer and Henk Visscher

OTB – Research for the Built Environment, Faculty of Architecture and the Built Environment, Delft University of Technology, Delft,
the Netherlands

ABSTRACT KEYWORDS
The difference between actual and calculated energy is called the ‘energy-performance gap’. big data; energy
Possible explanations for this gap are construction mistakes, improper adjusting of equipment, consumption; energy
excessive simplification in simulation models and occupant behaviour. Many researchers and epidemiology; energy
performance; household
governmental institutions think the occupant is the main cause of this gap. However, only
energy; occupant behaviour;
limited evidence exists for this. Therefore, an analysis is presented of actual and theoretical performance gap
energy consumption based on specific household types and building characteristics. Using a
large dataset (1.4 million social housing households), the average actual and theoretical energy
consumptions (gas and electricity) of different household types and characteristics (income level,
type of income, number of occupants and their age) were compared for each energy label.
Additionally, the 10% highest and lowest energy-consuming groups were analysed. The use of
combinations of occupant characteristics instead of individual occupant characteristics provides
new insights into the influence of the occupant on energy demand. For example, in contrast to
previous studies, low-income households consume more gas per m2 (space heating and hot
water) than households with a high income for all types of housing. Furthermore, the
performance gap is caused not only by the occupant but also by the assumed building
characteristics.

to set energy-saving targets and develop policies. For

Introduction
example, the Dutch social housing associations signed a
In 2002, the European Union introduced the Energy Per- covenant to renovate their building stock to reach an
formance of Building Directive (EPBD). The EPBD average energy label B by 2021, and thereby an energy
requires buildings to have an energy performance certifi- reduction of 33% between 2008 and 2021 (Bzk, Aedes,
cate (EPC), or energy label, when sold or rented. In the Woonbond, & Vastgoed Belang, 2012).
Netherlands, the energy label is calculated based on The discrepancies between actual (measured by
both the building characteristics and modelled heating energy distribution companies) and theoretical energy
behaviour of occupants. Through a simplified heat-trans- consumption (as calculated by the energy label) were
fer calculation, a theoretical energy usage is determined found by several researchers (Guerra-Santin & Itard,
that relates to an energy label. The theoretical energy 2010; Majcen, Itard, & Visscher, 2013b; Menezes, Cripps,
usage for residential buildings contains building-related Bouchlaghem, & Buswell, 2012; Sunikka-Blank & Galvin,
energy usage (e.g. energy for heating, hot water, venti- 2012). This set of discrepancies is known as the ‘energy-
lation, lighting in communal areas). Energy use for elec- performance gap’. Majcen et al. (2013b) showed that
trical appliances and lighting in private areas is excluded. occupants of ‘energy-inefficient’ buildings consume less
The aim of this energy label is to show potential buyers or gas (for space heating and hot water) than expected,
renters the energy efficiency of their dwelling in a simple while occupants of ‘energy-efficient’ buildings consume
and comprehensible way (Rijksoverheid, 2016a). Apart more than expected. Apart from gas, there is also a gap
from this, the labelling system is used by policy-makers between theoretical and actual electricity consumption.

CONTACT Paula van den Brom [email protected]

© 2017 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/
4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon in
any way.
2 P. VAN DEN BROM ET AL.

However, this gap for electricity is expected because occupancy consumption (theoretical consumption) are
theoretical energy consumption only incorporates build- essential for the development of energy-saving policy
ing-related energy consumption and not other electricity instruments (Bordass, Leaman, & Ruyssevelt, 2001;
consumption (e.g. electrical appliances and lighting). The Visscher, Meijer, Majcen, & Itard, 2016b). Also, most
performance gap for gas consumption is more difficult to studies that investigate actual energy consumption
explain because it primarily contains energy consump- focus either on occupant behaviour or the building’s
tion for heating, which is dependent on multiple factors. characteristics. The rebound effect, however, suggests
Several researchers found a significant influence of the an interaction between behaviour and building charac-
occupant on residential energy consumption (Gram-Hans- teristics. Understanding occupant behaviour is essential
sen, 2012; Palmborg, 1986; Sonderegger, 1978; Steemers & to predict the energy performance of buildings (Visscher,
Yun, 2009). Some even claim that the energy-performance Laubscher, & Chan, 2016a). Therefore, the present study
gap is primarily caused by occupant behaviour (Aydin, investigates the research question:
Kok, & Brounen, 2013; Gram-Hanssen, 2011). This suggests
Can analysing actual energy consumption by specific
that occupants in energy inefficient dwellings behave more household types and building characteristics contribute
energy efficiently than occupants in more energy-efficient to a better understanding of the role of the occupant in
dwellings. Additionally, occupants in energy-efficient dwell- actual energy consumption and the energy-performance
ings are assumed to have a higher comfort level than occu- gap?
pants in less energy-efficient dwellings, which could be an This research uses large databases. The first database
explanation for the underestimation of high energy-efficient is the SHAERE database from the umbrella organization
buildings. Guerra Santin (2013), for example, found that the of the Dutch social housing associations. It contains
average indoor temperature in energy-efficient dwellings is building characteristics and theoretical energy consump-
higher than in energy-inefficient dwellings. This can partly tion data from 1.4 million social rented houses in the
be explained by the so called ‘rebound effect’, which is Netherlands. The other two databases contain occupant
defined by Herring and Sorrell (2009) as the increase of characteristics and annual energy consumption data
energy consumption in services for which improvements from Statistics Netherland. By combining occupant
in energy efficiency reduce the costs.1 The opposite of the characteristics and analysis per energy label, it is possible
rebound effect is also found to be true, also known as the to use large databases to investigate the influence of the
‘pre-bound effect’ (Sunikka-Blank & Galvin, 2012). occupant on residential energy consumption (Hamilton
It is generally known that occupants influence resi- et al., 2013) and identify clear patterns and trends.
dential energy consumption. However, researchers This paper is structured as follows. The next section
have so far only been able to use occupant behaviour presents an overview of the literature on the influence
to explain some of the variance. For example, Guerra of occupant behaviour on residential energy consump-
Santin (2010) found evidence for 3.2–9.4% of the var- tion along with an explanation of the Dutch energy
iance in energy consumption due to occupant behaviour, label system. Then an overview of the databases and a
and Majcen (2016) for 9.1%. Despite limited evidence for description of the methods are provided. The findings
the actual influence of occupant behaviour on residential are then described. The final two sections contain the
energy consumption, several organizations and govern- discussion and conclusions.
ments have implemented campaigns to change energy
behaviours. A clear knowledge base of how inhabitants
actually use energy is necessary to improve the effective- Existing studies
ness of energy-saving campaigns to help policy-makers
set more realistic energy-saving targets, and to reduce This section describes findings of previous research
the energy-performance gap. However, it is rather regarding the influence of occupant behaviour on resi-
time-consuming and intrusive to gather actual occupant dential energy consumption. Findings that are not
behaviour data. As there is relatively little explanation for from the Netherlands are noted as such in the text.
the discrepancy in actual and theoretical energy use, bet-
ter insight into the influence of the occupant on residen-
Influence of actual behaviour on energy
tial energy consumption is required.
consumption
The lack of available occupant data is probably one of
the reasons researchers found only limited evidence for Residential energy includes energy for lighting and appli-
the influence of occupant behaviour on the performance ances, cooking, domestic hot water, and heating. In the
gap. However, results from in-use building performance Netherlands, heating consumes the largest share of a
research (actual energy consumption) instead of pre- building’s energy (Bosseboeuf, Gynther, Lapillonne, &
 BUILDING RESEARCH & INFORMATION 3

Pollier, 2015). It is widely recognized that building Occupant characteristic data are available on a larger
characteristics influence the actual energy consumption scale than occupant behaviour data. Additionally, cor-
in terms of heating. For example, buildings with a high relations between occupant characteristics and energy
insulation level consume less energy for heating than consumption are more usable for policy-makers than
buildings with a low insulation level. However, occupant actual behaviour data. Therefore, many researchers
behaviour is also found to have an effect on actual energy focus on occupant characteristics instead of actual
consumption for heating. For example, the hours that behaviour to study the influence of occupant behaviour
heating is at its maximum temperature explains 10.3% on residential energy consumption. The text below
of the variance in actual energy consumption for heating. describes the findings of previous research on the influ-
The number of hours the radiator is on in a certain room ence of occupant characteristics on gas and electricity
also explains a part of the variance of actual energy con- consumption.
sumption for heating (living room 8.8%, bedroom 8.1% Incomes in England were found to be positively cor-
and bathroom 5.9%) (Guerra Santin, 2011). related with the actual energy consumption in a house-
Furthermore, in China the set-point temperature was hold (Druckman & Jackson, 2008; Steemers & Yun,
found to influence significantly residential energy con- 2009). A 1% increase in income increases the total energy
sumption (O’Neill & Chen, 2002). Lowering the set- consumption by 0.63%, according to Vringer and
point temperature by 1°C can result in a significant Blok (1995). The correlation for electricity (r = 0.25;
reduction in energy use, similar to roof p < 0.01) was found to be marginally stronger than for
insulation (Guerra Santin, Itard, & Visscher, 2009). The gas (r = 0.23; p < 0.01) (Druckman & Jackson, 2008). A
set-point temperature at night and in the evening has larger number of household members also results in
more impact on total energy use than the temperature set- higher energy consumption, but it decreases the energy
ting during the day (Guerra Santin et al., 2009). consumption per person (Chen, Wang, & Steemers,
Appliances are the second main energy consumer in an 2013; Druckman & Jackson, 2008; Guerra Santin et al.,
average Dutch household (Bosseboeuf et al., 2015). 2009; Guerra-Santin & Itard, 2010; Jeeninga, Uyterlinde,
Research in the UK found that 19% of energy is consumed & Uitzinger, 2001; Kaza, 2010; O’Neill & Chen, 2002;
by stand-by and continuous appliances (e.g. refrigerators) Vringer & Blok, 1995; Yohanis, Mondol, Wright, & Nor-
(Firth, Lomas, Wright, & Wall, 2007). In Denmark, 10% ton, 2007; Yun & Steemers, 2011).
of household energy is used solely for stand-by appliances Age is found to be the most determining indirect
(Gram-Hanssen, Kofod, & Petersen, 2004). More frequent effect on heating and cooling energy use in different
use of electrical appliances over previous years has resulted countries (Guerra Santin, 2010; O’Neill & Chen, 2002;
in an increase of electricity consumption. For example, Pettersen, 1994; Yun & Steemers, 2011). Occupants
more frequent use of dishwashers has caused a decrease between 40 and 50 years demand the highest comfort
of gas consumption for hand washing but increased electri- and also have the highest average net income (Yohanis,
city use (Dril, Gerdes, Marbus, & Boelhouwer, 2012). 2011; Yohanis et al., 2007). Households with young chil-
Energy for domestic hot water is the third highest dren ventilate less, whereas households with older chil-
energy consumer in an average Dutch household (Bosse- dren ventilate more (Guerra Santin, 2010).
boeuf et al., 2015). The energy used for domestic hot Education level has only a very limited impact on resi-
water is, apart from the domestic hot water system, dential energy consumption. Higher-educated people set
strongly related to the number of people per household their thermostat for fewer hours on the highest tempera-
(Gerdes, Marbus, & Boelhouwer, 2014). The majority ture set-point than lower-educated people (Guerra
of domestic hot water is used for showering or bathing. Santin, 2010).
The frequency of showers has been stable in recent Household size and the presence of teenagers in the
years (on average 12 times a week per household) house is found to have a significant effect on energy con-
(Gerdes et al., 2014). sumption for appliances (Brounen, Kok, & Quigley,
Energy use for cooking has decreased in recent years. 2012).
People go out for dinner more often, and delivery and Finally tenants are found to have a higher rebound
takeaway meals are more common (Gerdes et al., 2014). effect than homeowners (tenants 31–49% and home-
owners 12–14%).
These results show that studying occupant character-
Influence of occupant characteristics on actual
istics is an effective way to investigate the influence of
energy consumption
occupants on residential energy consumption. Addition-
Several studies show a correlation between actual ally, studying occupant characteristics instead of actual
energy consumption and occupant characteristics. behaviour data enables one to work with larger datasets.
4 P. VAN DEN BROM ET AL.

Other explanations for the energy- calculation. The entire calculation and determination
performance gap method of the energy label can be found in ISSO 82.3
(2011) (energieprestatie advies woningen).
Although occupant behaviour is expected to be one of
As mentioned above, the theoretical energy is based
the main explanations for the energy-performance gap,
on a simplified heat-loss calculation. The air tightness,
other possible explanations should not be neglected.
insulation level and ventilation rate are taken into
The insulation level of the building is seldom measured;
account to define the energy demand for heating. The
in most cases it is estimated based on available building
energy consumption for domestic hot water is based
documents. As little or no data are available for older
on the assumed domestic hot water use in litres and
buildings, the insulation level of these buildings is deter-
the energy efficiency of the domestic hot water installa-
mined based on the construction year of the building.
tion. The theoretical energy consumption only contains
Recent research by Rasooli, Itard, and Infante Ferreira
building-related energy usage, which is the sum of pri-
(2016a) suggests that these assumptions could be an
mary energy for heating, domestic hot water, pumps/
important explanation for a part of the energy-perform-
fans and lighting in common areas minus the energy
ance gap.
gained from solar panels and cogeneration. This is also
Several studies show that the thermal mass of a building
important to consider when actual and theoretical
contributes significantly to its heating energy demand.
energy consumption are compared. The theoretical
This could be another explanation for the performance
energy consumption is calculated for a standard situ-
gap (Aste, Angelotti, & Buzzetti, 2009; Bojić & Loveday,
ation that assumes the following:
1997). However, the thermal mass is not taken into
account in the theoretical energy calculation of the . average indoor temperature of 18°C
Dutch EPC. Therefore, this could influence the discre- . average internal heat production due to appliances
pancy between actual and theoretical energy consumption. and people of 6 W/m2
Additionally, the theoretical energy-consumption cal- . 2620 degree-days (= 212 heating-days with an average
culation method used for the determination of the outdoor temperature of 5.64°C)
energy label only contains building-related energy con- . heating gains from the sun, vertical south orientation:
sumption. However, the actual energy consumption 855 MJ/m2
data also include occupant-related energy consumption . ventilation rate based on floor area and type of venti-
(e.g. use of electrical appliances). lation system
Finally, the theoretical energy consumption is calcu- . standard number of occupants based on floor area
lated with a steady-state model in this research. This (Table 1)
model might be oversimplified. The most oversimplified . 0.61 showers per day per person
aspects are assumed to be heat transfer between adjacent . 0.096 baths per day per person (if a bath is present).
rooms with an identical air temperature, the definition of
the combined radiative–convective heat-transfer coeffi- Qtotal = Qspace heating + Qwaterheating + Qaux.energy
cient, different definitions of solar gains (by surfaces or + Qlighting − Q pv − Qcogeneration (1)
the air), and including/excluding solar gains by exterior
surfaces such as roofs (Rasooli, Itard, & Infante Ferreira, where Qtotal = total theoretical primary energy con-
2016b). Time is not taken into account in the steady- sumption (MJ); Qspace heating = total theoretical pri-
state method, so the occupant behaviour is static in the mary energy consumption for space heating (MJ);
Dutch energy label calculation. However, the relationships Qwaterheating = total theoretical primary energy con-
between behaviour patterns and occupant characteristics sumption for domestic hot water (MJ); Qaux.energy =
are found in previous research (Kane, Firth, & Lomas, total theoretical primary energy consumption for
2015). The use of occupancy pattern models has signifi- pumps/ventilators (MJ); Qlighting = total theoretical
cantly improved the accuracy of the estimation in space-
heating energy use (Cheng & Steemers, 2011). Table 1. Assumed number of occupants in the theoretical energy
calculation (ISSO 82.3).
Floor area (m2) Assumed occupants
Dutch energy label < 50 1.4
50–75 2.2
This section describes briefly how the theoretical energy
75–100 2.8
consumption for Dutch dwellings is calculated and the
100–150 3.0
energy label is determined. Additionally, it describes
> 150 3.2
the assumptions made about the occupant in this
 BUILDING RESEARCH & INFORMATION 5

primary energy consumption for lighting (MJ); Qpv = occupants’ privacy. Furthermore, the data could only
total theoretical primary energy gains from solar (MJ); be accessed via a secured server from the CBS. The
and Qcogeneration = total theoretical primary energy data can only be exported on an aggregated level of at
gains from cogeneration (MJ). least 10 households.

Cleaning data
Data
The raw dataset was filtered before the analysis. First
This section describes the data used for this research and duplicate cases and cases that were not checked in
its representativeness. 2014 were removed from the dataset (reduction of
240,330 cases). Next, unrealistic floor areas for social
SHAERE database housing in the Netherlands (all dwellings smaller than
15 m2 and larger than 300 m2) were deleted (reduction
The SHAERE (Sociale Huursector Audit en Evaluatie van of 20,734 cases). Also, all cases with a gas-powered heat-
Resultaten Energiebesparing; in English, social rental sec- ing system that had a gas consumption of zero were
tor audit and evaluation of energy saving results) database removed, as were the cases with an electricity consump-
is owned by AEDES (the umbrella organization for Dutch tion of zero. Finally, all cases with a primary energy use
housing). Dutch social housing organizations own 31% of above 4000 MJ/m2 were deleted. The final dataset con-
the total housing stock in the Netherlands. The SHAERE tained 1,431,019 cases. A correction for climate was
database contains 60% of social housing stock. Besides applied though the application of degree-days. As the
building characteristics (e.g. insulation, type of glazing, energy consumption data of district and block heating
ventilation, heating and domestic hot water systems) the were found to be unreliable, all cases with this type of
SHAERE database also contains a pre-label and the corre- heating system were removed from the dataset.
sponding theoretical energy consumption and energy
index. A pre-label is a label that has not been validated
by the authorities but contains the same information as Household types
the validated one. The advantage of the pre-labels is that Based on occupant characteristic data, 18 household
they are made as soon as the energy performance of a types were formed. These are based on income, house-
house is upgraded. The database is updated every year. hold composition, type of income and age. These house-
For this research, the 2014 SHAERE database was used. hold types represent almost 80% of the total number of
cases in the SHAERE database (Table 2).
Household types are not equally distributed among
CBS (Statistics Netherlands Bureau) data
energy labels. Single households and retired couples
The theoretical energy consumption per dwelling is appear to live more often in A- and B-label dwellings
included in the SHAERE database, but to identify the than in the less energy-efficient dwelling types. Single
performance gap the actual energy consumption is households that receive state benefits or have a low to
required. For this research, the authors had access to average income live more often in dwellings with a low
the actual annual energy consumption data of Dutch energy label. The same applies for couples with a low
households provided by energy companies via the Stat- or average income and for receivers of state benefits.
istics Netherlands Bureau (CBS). This database contains Households with a high income on average live more
annual actual energy consumption at a household level. often in dwellings with a high energy label. Families
In addition, access was granted to occupant character- with children and a high income live more often in
istics data at a household level from the same source. dwellings with an energy label A. Families with a low
The occupant characteristics data include income, type or average income live less often in buildings with a
of income (from work, benefits etc.), household compo- high energy label (A and B) but also less often in build-
sition, number of occupants, occupants above and below ings with a low energy label (F and G).
age 65 years, number of children, and age of children.
This granularity of the data was available at the house-
Representativeness of the dataset
hold level. This allowed the research team to link those
databases and execute the analysis. This section compares the SHAERE database with the
This is one of the first studies to have access to such a national situation. First, the database contains only ren-
large and extensive database. Addresses and other per- tal dwellings data, which represents 55.8% of the housing
sonally identifiable data were encrypted to ensure the stock (Rijksoverheid, 2016b).
6 P. VAN DEN BROM ET AL.

Table 2. Household types.

Household composition Age (years) Children Age of children (years) Work Income
1 Single ≥65 No n.a Retired n.a
2 Single <65 No n.a State benefit n.a
3 Single <65 No n.a Employed Low
4 Single <65 No n.a Employed Middle
5 Single <65 No n.a Employed High
6 Couple >65 No n.a Retired n.a
7 Couple <65 No n.a State benefit n.a
8 Couple <65 No n.a Employed Low
9 Couple <65 No n.a Employed Middle
10 Couple <65 No n.a Employed High
11 Family <65 Yes < 12 State benefit n.a
12 Family <65 Yes < 12 Employed Low
13 Family <65 Yes < 12 Employed Middle
14 Family <65 Yes < 12 Employed high
15 Family <65 Yes At least one > 12 State benefit n.a
16 Family <65 Yes At least one > 12 Employed Low
17 Family <65 Yes At least one > 12 Employed Middle
18 Family <65 Yes At least one > 12 Employed High
Note: n.a. = Not applicable.

Compared with the national housing stock in the variations in floor area. Two methods are used. First,
Netherlands, the database contains fewer dwellings the theoretical and actual average energy consumption
with an energy label A and B (RVO, 2014). Compared for each household type per energy label are compared.
with the national housing stock, it contains more multi- The comparison is made on the energy label for two
family dwellings. Fewer buildings were constructed reasons. First, previous research found a relationship
before 1965 and between 1992 and 2005 in the database between occupant behaviour and the energy efficiency
than in the total national stock. More buildings were of the dwelling (Aydin et al., 2013; Sunikka-Blank & Gal-
constructed between 1965 and 1991 in the database com- vin, 2012). Second, the data revealed that household
pared with the national stock. types are unevenly distributed among the energy labels.
The average number of household members in SHAERE The statistical significance of this comparison is checked
(1.85) is lower than the overall national average in the with a linear regression.
Netherlands (2.20). A comparison between the assumed The second method is a more in-depth analysis of the
number of occupants in the energy performance calculation highest 10% and the lowest 10% energy-consuming
and that of the SHAERE database shows that the assumed groups of every energy label (Table 3). This approach
number is always higher than the actual number. was used because it is expected that the most relevant
The average income of the occupants in the database factors will be more clearly visible in the extreme groups
is lower than the average income of the total Dutch hous- than in the average group, where the factors will be less
ing stock. The first to the fifth income percentiles are visible because there is more noise. The assumption is
overrepresented and the higher income percentiles,
sixth to 10th, are underrepresented in the database.
Occupants over 65 years occur more often in Table 3. Households in the 10% high and lowest energy-
consuming groups.
SHAERE than in the national database (28.9% SHAERE,
Energy label Households
15% Dutch population). Particularly in dwellings with a
A 5018
better energy label, the number of people aged 65 years
B 18,076
and older is higher in the SHAERE database.
C 30,703
D 22,003
Method E 11,413
2 F 6330
Gas and electricity consumption per m are now studied.
This metric was chosen to reduce the impact of G 2442
 BUILDING RESEARCH & INFORMATION 7

that the observation of the extreme groups will dis- provides a better insight into their influence. Figures 2
tinguish the relevant parameters more quickly. Both and 3 show the results of this comparison. To keep the
groups are analysed for household type and other occu- results section concise, only results for energy labels B
pant characteristics as well as the building characteristics. and E are shown. The comparison results suggest that
The significance of the results is checked with a chi- actual energy consumption is more influenced by
square analysis. Analyses are conducted with IBM SPSS household type than theoretical energy consumption.
statistics 22 software. This is as expected because type of household is not
taken into account in the theoretical energy calculation
method.
Results
Single households have the lowest and family house-
The results are divided into two parts: gas consumption holds the highest gas consumptions for every energy
and electricity consumption. For both, first the difference label. This confirms previous research that a higher num-
between actual and theoretical consumption is explained ber of occupants results in higher gas consumption. Single
and then the highest and lowest energy-consuming and family households with a high income consume less
groups are compared. When interpreting the results, it gas in almost all cases compared with single and family
should be noted that the majority of the residential build- households with a low income for every energy label.
ings in the Netherlands (as in this database) use gas for These findings are confirmed by the regression analysis
space heating and domestic hot water. (Table 4) for the majority of household types. This contra-
dicts the findings of Vringer and Blok (1995). A possible
explanation is the use of gas consumption per m2 instead
Gas consumption
of total gas consumption.
Comparing actual and theoretical gas consumption per It is expected that people with a high income live in
energy label reveals that supposedly energy-efficient houses with a larger area, which they do not heat con-
buildings (energy labels A–B) consume more gas than stantly. However, if the same regression analysis is per-
expected. Buildings that are supposed to be inefficient formed with the floor area of the dwelling, then a
(energy labels C–G) consume less gas than expected negative relationship exists between income and gas
(Figure 1). These findings confirm the findings of Maj- consumption, although the impact is smaller
cen et al. (2013b). (Table 4). This suggests that the size of the floor
Household types are then added to the comparison area is only part of the explanation for why households
between actual and theoretical gas consumption. This with a high income are often in the low gas-

Figure 1. Comparison of actual versus theoretical gas consumption.

8 P. VAN DEN BROM ET AL.

Figure 2. Comparison of mean actual versus theoretical gas consumption per household group – energy label B.

Figure 3. Comparison of mean actual versus theoretical gas consumption per household group – energy label E.
 BUILDING RESEARCH & INFORMATION 9

Table 4. Comparison regression analysis of gas-consumption energy (reference dummy variable = Single high income).
Energy label B + area, Energy label E, Energy label E + area,
Energy label B, R 2 = 0.011 R 2 = 0.082 R 2 = 0.010 R 2 = 0.089
Standardized Standardized Standardized Standardized
Household group B B p B B p B B p B B p
Constant 11.73 0.00 18.29 0.00 16.26 0.00 24.58 0.00
Single. 65+. Retired –0.63 –17.67 0.00 –1.39 –0.10 0.00 –0.04 0.00 0.51 –0.24 –0.01 0.00
Single. State benefits –0.01 –0.31 0.75 –1.24 –0.07 0.00 –0.59 –0.03 0.00 –1.78 –0.08 0.00
Single. Low income –0.16 –0.01 0.01 –1.54 –0.06 0.00 –0.63 –0.02 0.00 –1.78 –0.05 0.00
Single. Middle income –1.36 –0.07 0.00 –2.43 –0.12 0.00 –1.82 –0.08 0.00 –2.76 –0.11 0.00
Couple. 65+. Retired –0.95 –0.06 0.00 –1.02 –0.06 0.00 –0.73 –0.03 0.00 –0.07 0.00 0.25
Couple. State benefits 0.04 0.00 0.69 –0.07 0.00 0.44 0.07 0.00 0.59 0.10 0.00 0.42
Couple. Low income 1.01 0.01 0.00 0.10 0.00 0.60 –0.32 0.00 0.23 –1.17 –0.01 0.00
Couple. Middle income –0.46 –7.09 0.00 –0.69 –0.02 0.00 –0.61 –0.02 0.00 –0.68 –0.02 0.00
Couple. High income –1.21 –14.23 0.00 –1.07 –0.03 0.00 –1.42 –0.03 0.00 –1.18 –0.03 0.00
Family. Children < 12. 1.37 0.04 0.00 1.21 0.03 0.00 1.22 0.03 0.00 0.88 0.02 0.00
State benefits
Family. Children < 12. 1.45 0.01 0.00 1.26 0.01 0.00 1.63 0.01 0.00 1.11 0.01 0.00
Low income
Family. Children < 12. 0.08 0.00 0.34 0.43 0.01 0.00 0.05 0.00 0.63 0.24 0.01 0.03
Middle income
Family. Children < 12. –0.85 –0.01 0.00 –0.07 0.00 0.64 –0.49 –0.01 0.02 0.04 0.00 0.83
High income
Family. One or more –0.98 0.03 0.00 1.45 0.05 0.00 1.08 0.03 0.00 1.49 0.05 0.00
children > 12. State
benefits
Family. One or more 1.52 0.01 0.00 1.91 0.01 0.00 1.19 0.01 0.06 1.83 0.01 0.00
children > 12. Low
income
Family. One or more 0.13 0.00 0.30 0.86 0.02 0.00 0.71 0.01 0.00 1.24 0.02 0.00
children > 12. Middle
income
Family. One or more –0.54 –0.01 0.00 0.49 0.01 0.00 –0.19 0.00 0.49 0.95 0.01 0.00
children > 12. High
income
Floor area –0.08 –0.28 0.00 –0.10 –0.29 0.00

consumption group than households with a low between the performance gap and occupant character-
income. Another possible explanation is that house- istics or there are other factors that have a higher influ-
holds with a high income may spend less time at ence on the performance gap. Another explanation is
home than households with a low income and, there- that the average household type behaviour is dependent
fore, consume less gas. on the energy efficiency of the dwelling, e.g. household
As expected, only a limited relationship was found types behave more energy efficiently in energy-ineffi-
between household type and theoretical energy con- cient than in energy-efficient dwellings (the pre-bound
sumption. The relationship can be traced back to house- effect).
hold characteristics.
The largest difference between average actual and
Highest and lowest gas-consuming groups
theoretical gas consumption in the total sample is
compared with the average
found for single households that receive state benefits.
The smallest difference is found for families with a To get a better insight into the actual energy consump-
high income from work. Analyses that take the energy tion, the households with the 10% highest and 10% low-
labels into account show the smallest performance gap est actual gas consumptions per energy label are
for family households in dwellings with a low energy analysed. The chi-square was used to test the statistical
label (D–G). Single households show the smallest gap difference in the distribution of the three groups (10%
for dwellings with an energy label between A and highest energy consumers, 10% lowest energy consumers
C. This means that there is no direct relationship and 80% average energy consumers).
10 P. VAN DEN BROM ET AL.

Figures 4 and 5 show the average actual and theoreti- consuming group than in the other groups, independent
cal gas consumption per energy label, the mean lowest of label type. With the exception of the single retired
10% and the mean highest 10% gas-consuming group. household, this group occurs more often in the lower
The difference between the highest, lowest and total gas-consumption group for labels A–C, and more
theoretical gas-consuming groups provides evidence often in the higher gas-consuming group for labels
that building characteristics influence actual energy con- F and G. This implies that the building characteristics
sumption. However, these differences are smaller com- have a larger influence on elderly people than on
pared with actual energy gas-consuming groups. This other household types. An explanation for this phenom-
suggests that other factors also influence the actual enon could be that elderly people are more often at
energy consumption. home and, therefore, heat their house for longer. How-
A comparison between the average actual and theoreti- ever, this explanation cannot be confirmed by this
cal gas consumptions for the lowest 10% gas-consuming research because actual occupant behaviour is not avail-
group shows an almost flat gas use for the actual gas con- able. The comparison also shows that family households
sumption and (as expected) an increasing theoretical with children aged 12 years and above occur more often
energy use as the label increases. The comparison of the in the higher gas-consuming groups for every label type.
actual and average theoretical gas consumption for Specific occupant characteristics were also compared.
the highest 10% gas-consuming group shows that even In agreement with previous studies, the number of
the average highest actual gas-consuming group con- household members shows that households with one
sumes less gas than the predicted actual gas consumption. member occur more often in the lower gas-consumption
To understand why residential buildings belong in the group, and households with three or more members
highest or lowest actual energy-consuming group, a more occur more often in the higher gas-consumption
detailed comparison was made. This involved the com- group. The difference in distribution is significant
parison of the highest and lowest energy-consuming (energy label B χ 2(8, N = 185,390) = 1832, p < 0.001 and
groups for both the building and occupant characteristics. energy label E χ 2(8, N = 115,659) = 1037, p < 0.001).
A comparison of the distribution of household types Households without children occur more frequently
for the total, highest and lowest gas-consuming groups in the low gas-consuming group and an increased num-
per energy label shows that the distribution of house- ber of children causes the household to occur more often
hold types is different between groups (energy label B in the higher gas-consuming group (Figure 6). The dis-
χ 2(34, N = 185,390) = 3747, p < 0.001 and energy label tribution difference between groups is significant (energy
E χ 2(34, N = 115,659) = 2287, p < 0.001) Single house- label B χ 2(8, N = 185,390) = 921, p < 0.001 and energy
holds occur more frequently in the lower gas- label E χ 2(12, N = 115,659) = 491, p < 0.001).

Figure 4. Comparison of highest, average and lowest mean theoretical gas consumption per energy label.
 BUILDING RESEARCH & INFORMATION 11

Figure 5. Comparison of highest, average and lowest mean actual gas consumption per energy label.

The chi-square test showed a significant difference of households occur more often in the average group. In
the distribution of household incomes between the high, the previous comparison per occupant group, however,
low and average energy-consuming groups (energy label we found that higher incomes are related to lower gas con-
B χ 2(18, N = 185,390) = 1332, p < 0.001 and energy label sumption. A possible explanation is the household type
E χ 2(18, N = 115,659) = 838, p < 0.001) Lower-income was not taken into account in this comparison. Therefore,
households occur more often in the extreme groups other household characteristics (e.g. number of house-
(high and low gas consumption) and higher-income hold members) can therefore influence the results.

Figure 6. Comparison of the distribution of the number of children in a household for the highest, average and lowest gas-consuming
group.
12 P. VAN DEN BROM ET AL.

If there is at least one household member who is buildings with a central heating system, a suggestion pre-
employed, the chance that this household belongs to the viously made by Majcen, Itard, and Visscher (2013a).
low energy-consuming group is higher than when no The distribution of housing type among the highest,
member is employed (energy label B χ 2(10, N = 185,390) lowest and average gas-consuming groups is also signifi-
= 430, p < 0.001 and energy label E χ 2(10, N = 115,659) cantly different (energy label B χ 2(16, N = 185,390) =
= 256, p < 0.001). A possible explanation for this could 4702, p < 0.001 and energy label E χ 2(16, N = 115,659)
be that the house is occupied fewer hours per day if some- = 2650, p < 0.001). Single-family houses occur more
one works. Also other studies found that occupation time often in the high-consuming groups, while apartments
influences residential energy consumption (Guerra Santin occur more often in the low gas-consuming groups.
et al., 2009; Majcen, Itard, & Visscher, 2015). This can partly be explained by single-family houses hav-
Apart from the occupant characteristics, Figure 4 ing a larger building envelope than apartments.
suggests that building characteristics also influence As expected, buildings that are well insulated (Rc >
whether a building belongs to the highest or the lowest 3.86) occur more often in the low-consuming group
gas-consuming group. Therefore, the distribution of cer- and buildings with poor or no insulation (Rc < 2.86)
tain building characteristics in the highest and lowest occur more often in the high-consuming group (energy
energy-consuming group are analysed per energy label label B χ 2(10, N = 185,390) = 2761, energy label E χ 2(8,
group. The influence of heating systems could only be N = 115,659) = 164). The results for energy label G
studied with some reservations because the condensing were not conclusive. The average U-value of windows
boiler is present in more than 90% of A–D dwellings; is lower for the low energy-consuming groups (energy
F and G dwellings have a higher mix of heating systems. label B χ 2(10, N = 185,390) = 630 and energy label B
Analysing the heating systems shows that the gas fire (an χ 2(10, N = 115,659) = 197).
appliance that heats an individual room) occurs Mechanical exhaust ventilation and natural ventilation
more frequently in the low energy-consuming group, occur more often in the high energy-consumption group
despite a low energy-efficiency rating (energy label from label A (Figure 7), while a balanced ventilation sys-
B χ 2(12, N = 185,390) = 213, p < 0.001 and energy label tem occurs more often in the low energy-consumption
E χ 2(14, N = 115,659) = 712, p < 0.001). A possible expla- group (energy label A χ 2(6, N = 185,390) = 2132, p <
nation is that gas fires cannot heat the same floor area as 0.001, energy label B χ 2(9, N = 192,354) = 6779, p < 0.001

Figure 7. Comparison of the distribution of ventilation systems for the highest, average and lowest gas-consuming group.
 BUILDING RESEARCH & INFORMATION 13

and energy label C χ 2(6, N = 115,659) = 356, p < 0.001). Electricity

Labels B and C have a negligible number of balanced ven-
Comparisons of the average actual and theoretical elec-
tilation systems; therefore, mechanical exhaust ventilation
tricity consumption per household type divided per
occurs more often in the low energy-consuming group and
energy label show a difference among household
natural ventilation in the high energy-consuming group.
types (Figures 9 and 10). Single households consume
No conclusive results were found for the buildings with
the least electricity per m2 of floor area. Families,
an energy label lower than C because they have a lower
especially those with children above 12 years of age,
variety in ventilation systems.
consume the most energy. Families that receive state
Within the A label group, older buildings occur more
benefits have a lower electricity consumption than
often in the high gas-consuming group than newer build-
people who have a high income from work. For
ings (Figure 8). It is highly unlikely that buildings built
couples, the electricity consumption for people with
before 1991 had an energy label A from origin, because
state benefits is a little higher than for employed
building regulations did not require it. It is expected,
people. Couples with a low income consume relatively
therefore, that the buildings with an older construction
the least electricity.
year in label-A dwellings are renovated. Our findings
suggest that it is difficult for renovated buildings to
reach the same energy-performance level as newer build-
Highest and lowest electricity-consuming group
ings. Fuel poverty could be another explanation. How-
compared with the average
ever, it is less probable because we found the amount of
high-income households in this group is five times higher The 10% highest and 10% lowest electricity consumer
than the amount of low-income households. groups were analysed for electricity consumption. Little
These findings support the general idea that the input difference was found for the influence of household
for theoretical energy calculations for buildings with a type per energy label. As a consequence, energy labels
high energy label is more reliable than the input for are not taken into account in this analysis. The distri-
buildings with a low energy label. More assumptions bution of household types between the high, low and
are likely made about the input for older buildings average energy-consuming groups differs significantly
than for more recent buildings due to the availability (χ 2(34, N = 1,100,756) = 55,441, p < 0.001). Single-occu-
of data. pant households occur more often in the lower-income

Figure 8. Comparison distribution of construction year for the highest, average and lowest gas-consuming group.
14 P. VAN DEN BROM ET AL.

Figure 9. Comparison of mean actual versus theoretical electricity consumption per household group – energy label B.

Figure 10. Comparison of mean actual versus theoretical electricity consumption per household group – energy label E.
 BUILDING RESEARCH & INFORMATION 15

groups than families. Single retired households occur energy companies are only required to report energy
most frequently in the lowest electricity-consuming consumption every three years. This means that the
group. In the higher electricity-consuming groups, data provided are not necessarily the actual data from
families and couples occur most frequently, especially 2014, but more likely to be data from 2012 or 2013.
families with children older than 12 years. Although this is a serious limitation of the dataset, this
Occupants with a high income occur more frequently is the best available data on such a large scale.
in the higher electricity-consuming groups. Occupants The theoretical energy calculation method is only a
with a low income occur more frequently in the low elec- simplified version of reality. Therefore, it is not realistic
tricity-consuming groups (χ 2(18, N = 1,100,756) = to expect it to bridge the energy-performance gap at the
15,126, p < 0.001). level of individual households. However, it should be
Also, a significant difference was found for the distri- able to reduce the gap for average energy consumption.
bution of the number of people per household (χ 2(8, N = For this reason, this research focused mainly on average
1,100,756) = 42,472, p < 0.001). Single households occur energy consumption. Although general conclusions can
more often in the low energy-consuming group and be drawn for specific socio-economic household types,
households with two or more members occur more it should be noted that each household is unique and,
often in the high energy-consuming groups. therefore, the occupants’ behaviour can be different
from the average.
The occupant characteristics data used in this
Discussion
research do not account for changes within household
One of the strengths of this study is the extensive dataset, demographics during the year, e.g. domestic separations,
with 1.4 million dwellings. In contrast to most studies of the birth of children and becoming unemployed.
occupants’ influence on residential energy consumption, Despite these limitations, this research provides new
the present study takes both occupant characteristics and insights into the influence of occupant characteristics
building characteristics into account. This sample only on actual energy consumption and provides several indi-
contains buildings owned by social housing organiz- cations for further research.
ations in the Netherlands, therefore all dwellings are ren-
tal dwellings. Studies in Germany and the Netherlands
Conclusions
show that tenants behave differently from housing own-
ers. For example, the rebound effect for tenants is found The findings of this research show that analysing specific
to be significantly larger than for homeowners (Aydin household types and building characteristics contributes
et al., 2013; Madlener & Hauertmann, 2011). to a better understanding of the influence of the occu-
The main target group of Dutch housing associations pant on actual energy consumption and the energy-per-
are people with a low income and, therefore, the average formance gap. The analysis of the highest and lowest
income of the sample size is lower than that of the entire 10% of consumers can help policy-makers to choose
Dutch population. Additionally, the average number of the right target groups for their energy-saving policies
household members is relatively low in the SHAERE and campaigns. Energy-saving advice can also be tailored
database compared with the national average. This may to specific household types.
have influenced our findings. The results imply that the building characteristics
In the data-filtering process, several possible mistakes have a higher impact on the elderly than on younger
were found in both the SHAERE data and CBS datasets. people. This could be an incentive for policy-makers to
Although the current authors tried to reduce the amount prioritize building renovations for the elderly.
of incorrect data as much as possible, there could still be Single households with a high income are found to
cases with incorrect data. Remaining sources of errors have the lowest average energy consumers. A possible
could be due to mistakes in the technical process, such explanation could be that they spend less time at home
as meter uncertainties, or translation mistakes from compared with other household types. Therefore,
one database to the other, and human mistakes during energy-saving campaigns focused on residential behav-
the registering process of the houses in the SHAERE iour might be not the most effective strategy. However,
database, which is performed manually. families with a low income or those that receive state
Housing organizations ought to update their data- benefits could benefit from energy-saving campaigns
bases each year, but it is not known how accurately or focusing on the reduction of gas consumption. For the
in how much detail they update the state of their building reduction of electricity consumption, this research
stock. Also, the accuracy of the actual energy consump- suggests that focusing on families with high incomes
tion data from the CBS is not known. Additionally, would be most effective.
16 P. VAN DEN BROM ET AL.

The analysis reveals a disparity between buildings in performance of well insulated buildings. Energy and
the same energy group. Buildings constructed more Buildings, 41, 1181–1187. doi:10.1016/j.enbuild.2009.06.005
recently consume less energy than older buildings within Aydin, E., Kok, N., & Brounen, D. (2013). The rebound effect in
residential heating. Retrieved September 29, 2016, from
the same energy-label grouping. The energy performance https://www.tilburguniversity.edu/upload/5301e4d6-0312-405
of a new building with energy label A is not the same as a 4-b680-cfd88a1525f7_The20Rebound20Effect_EA300813.pdf.
renovated building with an energy label A. This suggests Bojić, M. L., & Loveday, D. L. (1997). The influence on build-
that although renovated buildings reach similar energy ing thermal behavior of the insulation/masonry distribution
performances on paper, these are not achieved in prac- in a three-layered construction. Energy and Buildings, 26,
153–157. doi:10.1016/S0378-7788(96)01029-8
tice. A consequence is that expectations (and financial
Bordass, B., Leaman, A., & Ruyssevelt, P. (2001). Assessing
and other formulations) will need to be different in building performance in use 5: Conclusions and impli-
order to reflect this reality. cations. Building Research & Information, 29, 144–157.
The results of this research could also be beneficial for doi:10.1080/09613210010008054
energy consultants and authorities responsible for pro- Bosseboeuf, D., Gynther, L., Lapillonne, B., & Pollier, K. (2015).
viding EPCs. Additionally, the findings can help consult- Energy efficiency trends and policies in the household and
tertiary sectors; an analysis based on the ODYSSEE and
ants to explain to their clients that energy consumption MURE databases. ODYSSEE-MURE project. Retrieved
is not only dependent on physical factors but also on from http://www.odyssee-mure.eu/publications/br/energy-
occupants’ behaviour. efficiency-trends-policies-buildings.pdf: Intelligent Energy
Although a reduction of the performance gap was not a Europe Programme of the European Union.
goal, the findings can be used to interpret better the results Brounen, D., Kok, N., & Quigley, J. M. (2012). Residential
energy use and conservation: Economics and demographics.
of energy simulation. People who make building simu-
European Economic Review, 56, 931–945. doi:10.1016/j.
lations can, for example, inform their clients about the euroecorev.2012.02.007
differences between actual and theoretical energy con- Bzk, M. V. B. Z. E. K., Aedes, V. V. W., Woonbond, N., &
sumption and the possible explanations for these differ- Vastgoed Belang, V. V. P. B. I. V. (2012). Convenant ener-
ences. This can help clients understand why actual giebesparing huursector.
energy consumption is sometimes higher than expected Chen, J., Wang, X., & Steemers, K. (2013). A statistical analysis
of a residential energy consumption survey study in
and thus prevent disappointment. Hangzhou, China. Energy and Buildings, 66, 193–202.
Nevertheless, more research is required. In this doi:10.1016/j.enbuild.2013.07.045
research, relationships between certain occupant charac- Cheng, V., & Steemers, K. (2011). Modelling domestic energy
teristics and actual energy consumption were found, but consumption at district scale: A tool to support national and
the causes of these relationships were not investigated. local energy policies. Environmental Modelling & Software,
26, 1186–1198. doi:10.1016/j.envsoft.2011.04.005
To explain these relationships, a similar study should
Dril, T. V., Gerdes, J., Marbus, S., & Boelhouwer, M. (2012).
be executed on more specific actual behaviour data. A Energie Trends 2012. Petten: ECN, Energie-Nederland en
smaller database should be sufficient for this follow-up Netbeheer Nederland.
research. Druckman, A., & Jackson, T. (2008). Household energy con-
sumption in the UK: A highly geographically and socio-
economically disaggregated model. Energy Policy, 36,
Note
3177–3192. doi:10.1016/j.enpol.2008.03.021
1. An example of the rebound effect is when a home is ret- Firth, S., Lomas, K., Wright, A., & Wall, R. (2007). Identifying
rofitted with insulation or a more efficient boiler. The trends in the use of domestic appliances from household
expected efficiency gain is negated if people increase electricity consumption measurements. Energy and
the hours of space heating and/or raise the internal (win- Buildings, 40, 926–936.
ter) temperature. This results in a higher energy use. Gerdes, J., Marbus, S., & Boelhouwer, M. (2014). Energie
Trends 2014. Petten: ECN, Energie-Nederland en
Netbeheer Nederland.
Disclosure statement Gram-Hanssen, K. (2011). Households’ energy use – which is
No potential conflict of interest was reported by the authors. the more important: Efficient technologies or user practices?
In EEE (Ed.), World renewable energy congress 2011 (pp.
992–999). Linkoping: Linköping University Electronic Press.
ORCID Gram-Hanssen, K. (2012). Efficient technologies or user
behaviour, which is the more important when reducing
Henk Visscher http://orcid.org/0000-0003-0929-1812 households’ energy consumption? Energy Efficiency, 2013,
447–457.
References Gram-Hanssen, K., Kofod, C., & Petersen, K. N. (2004).
Different everyday lives: Different patterns of electricity
Aste, N., Angelotti, A., & Buzzetti, M. (2009). The influence of use. In ACEEE summer study on energy efficiency in build-
the external walls thermal inertia on the energy ings (pp. 1–13). Sweden: EEE.
 BUILDING RESEARCH & INFORMATION 17

Guerra Santin, O. (2010). Actual energy consumption in dwell- domestic buildings: Using post-occupancy evaluation data
ings; the effect of energy performance regulations and occu- to reduce the performance gap. Applied Energy, 97, 355–
pant behaviour. TU Delft: Delft University Press. 364. doi:10.1016/j.apenergy.2011.11.075
Guerra Santin, O. (2011). Behavioural patterns and user profiles O’Neill, B. C., & Chen, B. S. (2002). Demographic determi-
related to energy consumption for heating. Energy and nants of household energy use in the United States.
Buildings, 43, 2662–2672. doi:10.1016/j.enbuild.2011.06.024 Population and Development Review, 28, 53–88.
Guerra Santin, O. (2013). Occupant behaviour in energy effi- Palmborg, C. (1986). Social habits and energy consumption
cient dwellings: Evidence of a rebound effect. Journal of in single-family homes. Energy, 11, 643–650. doi:10.1016/
Housing and the Built Environment, 28, 311–327. doi:10. 0360-5442(86)90144-1
1007/s10901-012-9297-2 Pettersen, T. D. (1994). Variation of energy consumption in
Guerra-Santin, O., & Itard, L. (2010). Occupants’ behaviour: dwellings due to climate, building and inhabitants. Energy
Determinants and effects on residential heating consump- and Buildings, 21, 209–218. doi:10.1016/0378-7788(94)
tion. Building Research & Information, 38, 318–338. 90036-1
doi:10.1080/09613211003661074 Rasooli, A., Itard, L., & Infante Ferreira, C. (2016a). A response
Guerra Santin, O., Itard, L., & Visscher, H. (2009). The effect of factor-based method for the rapid in-situ determination of
occupancy and building characteristics on energy use for space wall’s thermal resistance in existing buildings. Energy and
and water heating in Dutch residential stock. Energy and Buildings, 119, 51–61. doi:10.1016/j.enbuild.2016.03.009
Buildings, 41, 1223–1232. doi:10.1016/j.enbuild.2009.07.002 Rasooli, A., Itard, L., & Infante Ferreira, C. (2016b). Rapid,
Hamilton, I. G., Summerfield, A. J., Lowe, R., Ruyssevelt, P., transient, in-situ determination of wall’s thermal transmit-
Elwell, C. A., & Oreszczyn, T. (2013). Energy epidemiology: tance. REHVA European HVAC Journal, 53, 16–20.
A new approach to end-use energy demand research. Rijksoverheid. (2016a). Energielabel woningen en gebouwen
Building Research & Information, 41, 482–497. doi:10. [Online]. Retrieved from https://www.rijksoverheid.nl/
1080/09613218.2013.798142 onderwerpen/energielabel-woningen-en-gebouwen/inhoud
Herring, H., & Sorrell, S. (2009). Energy efficiency and sustain- /waarom-een-verplicht-energielabel.
able consumption. The rebound effect. Basingstoke: Palgrave Rijksoverheid. (2016b). Woningvoorraad naar eigendom,
Macmillan. 2006–2014 [Online]. Retrieved from http://www.clo.nl/
ISSO 82.3. (2011). ISSO 82.3 publication energy performance indicatoren/nl2164-woningvoorraad-naar-eigendom.
certificate - formula structure (Publicatie 82.3 Handleiding RVO. 2014. Compendium voor de leefomgeving [Online].
EPA-W (Formulestructuur). The Hague: SenterNovem. Retrieved from http://www.clo.nl/indicatoren/nl0556-ener
Jeeninga, H., Uyterlinde, M., & Uitzinger, J. (2001). gielabels-woningen.
Energieverbruik van energiezuinige woningen; effecten van Sonderegger, R. C. (1978). Movers and stayers: The resident’s
gedrag en besparingsmaatregelen op de spreiding in the hoogte contribution to variation across houses in energy consump-
van het reele energieverbruik. Petten: Energieonderzoek tion for space heating. Energy and Buildings, 1, 313–324.
Centrum Nederland. doi:10.1016/0378-7788(78)90011-7
Kane, T., Firth, S. K., & Lomas, K. J. (2015). How are UK Steemers, K., & Yun, G. Y. (2009). Household energy consump-
homes heated? A city-wide, socio-technical survey and tion: A study of the role of occupants. Building Research &
implications for energy modelling. Energy and Buildings, Information, 37, 625–637. doi:10.1080/09613210903186661
86, 817–832. doi:10.1016/j.enbuild.2014.10.011 Sunikka-Blank, M., & Galvin, R. (2012). Introducing the pre-
Kaza, N. (2010). Understanding the spectrum of residential bound effect: The gap between performance and actual
energy consumption: A quantile regression approach. energy consumption. Building Research & Information, 40,
Energy Policy, 38, 6574–6585. doi:10.1016/j.enpol.2010.06.028 260–273. doi:10.1080/09613218.2012.690952
Madlener, R., & Hauertmann, M. (2011). Rebound effects in Visscher, H., Laubscher, J., & Chan, E. (2016a). Building gov-
German residential heating: Do ownership and income ernance and climate change: Roles for regulation and related
matter? FCN Working paper, 2/2011. polices. Building Research & Information, 44, 461–467.
Majcen, D. (2016). Predicting energy consumption and savings doi:10.1080/09613218.2016.1182786
in the housing stock – A performance gap analysis in the Visscher, H., Meijer, F., Majcen, D., & Itard, L. (2016b).
Netherlands (PhD), Delft University of Technology. Improved governance for energy efficiency in housing.
Majcen, D., Itard, L., & Visscher, H. (2013a). Actual and theor- Building Research & Information, 44, 552–561. doi:10.
etical gas consumption in Dutch dwellings: What causes the 1080/09613218.2016.1180808
differences? Energy Policy, 61, 460–471. doi:10.1016/j.enpol. Vringer, K., & Blok, K. (1995). The direct and indirect energy
2013.06.018 requirements of households in the Netherlands. Energy
Majcen, D., Itard, L., & Visscher, H. (2015). Statistical model of Policy, 23, 893–910. doi:10.1016/0301-4215(95)00072-Q
the heating prediction gap in Dutch dwellings: Relative Yohanis, Y. G. (2011). Domestic energy use and householders’
importance of building, household and behavioural charac- energy behaviour. Energy Policy, 41, 654–665.
teristics. Energy and Buildings, 105, 43–59. doi:10.1016/j. Yohanis, Y. G., Mondol, J. D., Wright, A., & Norton, B. (2007).
enbuild.2015.07.009 Real-life energy use in the UK: How occupancy and dwell-
Majcen, D., Itard, L. C. M., & Visscher, H. (2013b). Theoretical ing characteristics affect domestic electricity use. Energy and
vs. actual energy consumption of labelled dwellings in the Buildings.
Netherlands: Discrepancies and policy implications. Yun, G. Y., & Steemers, K. (2011). Behavioural, physical and
Energy Policy, 54, 125–136. doi:10.1016/j.enpol.2012.11.008 socio-economic factors in household cooling energy con-
Menezes, A. C., Cripps, A., Bouchlaghem, D., & Buswell, R. sumption. Applied Energy, 88, 2191–2200. doi:10.1016/j.
(2012). Predicted vs. actual energy performance of non- apenergy.2011.01.010