Energy Economics 48 (2015) 145–156

Contents lists available at ScienceDirect

Energy Economics
journal homepage: www.elsevier.com/locate/eneco

Does energy efficiency matter to home-buyers? An investigation of EPC

ratings and transaction prices in England
Franz Fuerst a, Patrick McAllister b, Anupam Nanda c,⁎, Peter Wyatt c
a
University of Cambridge, UK
b
University College London, UK
c
University of Reading, UK

a r t i c l e i n f o a b s t r a c t

Article history: This paper investigates whether energy performance ratings, as measured by mandatory Energy Performance
Received 7 October 2013 Certificates (EPCs), are reflected in the sale prices of residential properties. This is the first large-scale empirical
Received in revised form 3 December 2014 study of this topic in England involving 333,095 dwellings sold at least twice in the period from 1995 to 2012.
Accepted 7 December 2014
Applying hedonic regression and an augmented repeat sales regression, we find a positive relationship between
Available online 3 January 2015
the energy efficiency rating of a dwelling and the transaction price per square metre. The price effects of superior
JEL Classification:
energy performance tend to be higher for terraced dwellings and flats compared to detached and semi-detached
R21 dwellings. The evidence is less clear-cut for rates of house price growth but remains supportive of a positive as-
R28 sociation. Overall, the results of this study suggest that energy efficiency labels have a measurable and significant
331 impact on house prices in England.
C21 © 2015 Elsevier B.V. All rights reserved.
C81
Q58

Keywords:
Energy efficiency
House price
Hedonic model
Repeat sales model

1. Introduction
As part of a broad objective of combating climate change, compulso-
ry energy labelling of residential and commercial real estate has been
introduced in many countries across the world. In the residential sector,
acting as a market-based environmental policy instrument, energy
labelling informs consumers about energy performance of dwellings.
The primary objective is to reduce uncertainty about quality, generate
positive demand shifts and, in turn, an upward shift in the price and
an increase in the supply of energy efficient dwellings. Given that they
provide the basis for the operation of well-functioning markets, prices
are a key mechanism for ‘harnessing market forces’ to achieve the objec-
tives of such market-based policies. Empirical research on the existence
and extent of the price effects of energy labelling is important in
evaluating the effectiveness of this type of policy instrument. This
paper reports on an investigation of the relationship between energy
performance ratings and sale prices of dwellings in England. Focussing
on repeat sales, details of transactions involving 333,095 dwellings
⁎ Corresponding author at: Henley Business School, University of Reading,
Whiteknights, Reading RG6 6UD, UK. Tel.: +44 118 378 6339; fax: +44 118 378 8172.
E-mail address: a.nanda@reading.ac.uk (A. Nanda).
that took place in the period from 1995 to 2012 have been analysed.
Before reviewing the data in greater detail, describing the statistical
approaches to their analysis and discussing the results, we first provide
some background and context to the role of energy labelling in the
English residential real estate market, review the literature on
eco-labelling and dwelling prices and specify a theoretical framework
that outlines the expected relationship between energy rating and price.
2. Energy labelling and real estate markets
Over the last decade, a wide range of energy monitoring instruments
and environmental labels has been introduced to both the commercial
and the residential sectors of the real estate market. Although these in-
struments are a mixture of mandatory and voluntary schemes, with
some highly energy efficient or zero carbon buildings collecting several
labels, the boundary between mandatory and voluntary environmental
labels has become blurred as urban planning authorities make labels
such as Code for Sustainable Homes, BREEAM and LEED a condition of
permission to develop. In 2008, the measurement of energy use in
new and existing buildings in the UK became obligatory as a result of
the European Union Energy Performance of Buildings Directive
(EPBD). The Directive required all buildings at the point of construction

http://dx.doi.org/10.1016/j.eneco.2014.12.012
0140-9883/© 2015 Elsevier B.V. All rights reserved.
146 F. Fuerst et al. / Energy Economics 48 (2015) 145–156
completion, sale or rent (or every ten years) to have certificates giving
information about their energy performance. For residential dwellings
in the UK, an Energy Performance Certificate (EPC), and the accompany-
ing report, is an asset rating which is intended to inform potential
buyers or occupiers about the intrinsic energy performance of a building
and its associated services as built. EPCs are similar to the mandatory
energy labels used in many consumer products such as tumble dryers
and washing machines; buildings are rated on a scale A–G with band
A being the most efficient.
In the 1980s, a body of work emerged that investigated the relation-
ship between energy efficiency (typically proxied by energy bills) and
residential sale prices. Laquatra et al. (2002) provide a useful evaluation
of this work. Among a range of limitations identified, probably the most
important is that the studies typically involve small, highly localised sam-
ples consisting of dozens or hundreds of dwellings. However, they tend to
find a positive relationship between energy efficiency (or proxies for en-
ergy efficiency) and residential sale prices (see Dinan and Miranowski,
1989; Halvorsen and Pollakowski, 1981; Johnson and Kaserman, 1983;
Laquatra, 1986; Quigley, 1984; Quigley and Rubinfeld, 1989).
As eco-labelling has become more prevalent, a more recent body of
research has emerged that focuses on the price effects of intrinsic poten-
tial energy efficiency as revealed by a certificate or label rather than
realised performance outcomes. To date, the majority of these studies
have concentrated on the US commercial office market. Broadly, the
findings reveal a positive relationship between environmental labels
and price (see Wiley et al., 2010; Eichholtz et al., 2010, 2013; Fuerst
and McAllister, 2011a,b; Reichardt et al., 2011; Deng et al., 2011). One
of the first investigations of the price effect of mandatory energy
labelling in a residential real estate market was carried out by the
Australian Bureau of Statistics (2008). Using a sample of residential
sales in the Australian Capital Territory for 2005 (2385 transactions)
and 2006 (2719 transactions) and standard hedonic procedures, the
Bureau estimates the effect of Energy Efficiency Rating (EER) on house
prices. For 2005, they estimate a premium of approximately 1% for
every 0.5 increase in EER rating (EER ranges from 0 to 5). For 2006,
they estimate a premium of approximately 2% for every 0.5 increase in
EER. For a pooled sample, relative to zero rating house, they estimate
premiums of 1.6% (EER 1), 3% (EER 2), 5.9% (EER 3), 6.3% (EER 4) and
6.1% (EER 5). The explanatory power of the models is high and a large
number of control variables for dwelling quality are included. It is
worth noting that they find evidence of a nonlinear effect — the marginal
addition to the price effect declines as rating increases.
For The Netherlands office market, Kok and Jennen (2012) looked at
the relationship between EPC rating and rental price for 1057 transac-
tions in the period 2005–2010. Using standard hedonic techniques,
they identified a rental premium of approximately 4.7% for buildings
rated C or better compared to buildings rated D and worse. However, it
is possible that offices rated C or above may be better quality than build-
ings with inferior performance. Put simply, the level of energy efficiency
may be correlated with unobserved quality variables such as design and
interior finish. In the UK, Fuerst and McAllister (2011c) examined effect
of EPC rating on capitalisation rate, (appraised) market value and market
rent for 708 commercial properties as at September 2010. They found no
significant effect of EPC rating on appraised market rent or market value
and weak evidence of an effect on capitalisation rates.
In one of the most closely related studies to this research, Brounen
and Kok (2011) examined the relationship between EPC rating and sale
price for 31,993 residential sales in 2008–2009 in The Netherlands. Com-
pared to D-rated homes, they estimate premiums of 10%, 5.5% and 2% for
A, B and C ratings respectively. For dwellings rated E, F and G, they iden-
tified discounts of 0.5%, 2.5% and 5% respectively. Their sample contained
a broad range of control variables including dwelling size, insulation
quality, central heating and level of maintenance. It is worth noting that
although EPCs are mandatory in The Netherlands, a policy regarding ex-
emptions effectively made them optional. Consequently the adoption
rate is low. For instance, in August 2009 this rate dropped to seven
percent and never exceeded 25% throughout the study period. The au-
thors account for the quasi-voluntary nature of the EPCs with a Heckman
correction for selection bias, albeit the explanatory power of their model
of label adoption is low.
Another closely related study was conducted by Hyland et al. (2013)
who applied a standard hedonic technique to investigate the impact of
energy efficiency ratings on capital and rental asking prices for dwellings
in Ireland. Because only 5% of homes for sale (and only 2.3% of homes for
rent) have an efficiency rating, the Heckman procedure was employed to
control for selection bias. The results showed that, for a sample of 15,060
dwellings on the market between 2008 and 2012, there was a 9.3% price
premium for A-rated compared to D-rated dwellings, 5.5% for a B-rating,
and a 10.6% discount for F and G ratings. For rented dwellings the premi-
um for an A-rating was 1.8%, 3.9% for a B rating, a discount of 1.9% for E
ratings and 3.2% for F and G ratings. These results are broadly in line
with the findings from England that are presented in this paper. A limi-
tation of this investigation is the lack of a control for age, which is likely
to be significantly negatively correlated with the energy efficiency rating.
By contrast, the study presented here includes and pools information on
age and vintage from two independent sources as detailed in the ‘Data
and descriptive statistics’ section of this paper.
A key distinction between our analysis and both the Brounen and
Kok (2011) and Hyland et al. (2013) papers is that EPCs are mandatory
for virtually all new-build and dwelling transfers in the UK with very
few exceptions. This means that selection bias due to missing EPCs, a
problem that affects comparable studies published to date, is not a rel-
evant concern here. A second important distinction is our use of official
price information as recorded by the authorities (Land Registry data)
compared to list prices (Hyland et al., 2013) or transaction price data
gathered by realtor organisations (Brounen and Kok, 2011).
In summary, whilst there are fairly plausible a priori grounds to
expect a willingness to pay for energy efficiency by housing consumers,
in contrast to the large volume of research on the effects of school qual-
ity, accessibility and other amenities on houses prices, the empirical
research on the effect of energy or environmental labelling remains lim-
ited. Most of the work is focussed on commercial real estate markets.
Nearly all studies apply a version of Rosen's (1974) hedonic model to
estimate the price effect of an environmental label. In this paper we
build on the work described above and use a large sample of repeat
sale transactions to investigate whether and to what extent an energy
rating might influence dwelling price and the rate of change in dwelling
price. Before describing the dataset and method, in the next section we
attempt to formalise a theoretical framework that may be used to ana-
lyse the effect of EPC ratings on property prices.
3. Theoretical framework
We employ a hedonic model to estimate the effect of EPC ratings on
house prices. The theoretical foundations and empirical framework of
the model have been developed through many works, see Rosen
(1974), Edlefsen (1981), Roback (1982), Bartik (1987), Epple (1987),
Blomquist et al. (1988), Bajari and Benkard (2005) and many others
(see review of literature by Malpezzi, 2003). Essentially, the capitalisation
process works through constrained utility maximisation by consumers.
In the first instance, we assume that housing is a differentiated
economic good characterised by a vector H of several physical and loca-
tional attributes. For illustration, the list of these attributes consists of:
dwelling age (a), size (f), number of rooms (r), locational desirability (L)
and EPC rating (E). We also assume that there is a variable indicating
quality of the housing unit, Q. Assume one undifferentiated numeraire
commodity, c. The utility function, having the usual continuity and con-
vexity properties, is then specified as follows. Note that the continuity
and convexity properties are not applicable for non-continuous variables.
U ¼ U ðH; cÞ ð1Þ
 F. Fuerst et al. / Energy Economics 48 (2015) 145–156
H ¼ Hða; f ; r; L; Q; EÞ:
Moreover, it is also conceivable that location, quality and EPC rating
may also be determined by a multitude of factors, as shown in Eqs. (3),
(4) and (5):
L ¼ Lðd; xÞ
where d is distance to the town centre; x is a vector of urban features
and amenities such as level of infrastructure, school quality, crime rate
and availability of public services among others.
Q ¼ Q ðz; k; g Þ
where z is building material quality; k is a vector of design features; g is a
vector of energy performance features such as lighting, insulation, water
heating and glazing.
E ¼ Eðe; Q; LÞ
where e represents energy consumption and L and Q are as defined in
Eqs. (3) and (4) respectively.
Our question of interest is: how much is a consumer willing to pay for
a house with a specific attributes, namely an EPC rating? Consumers' util-
ity depends on income level and will be reflected by the revealed prefer-
ences of different types of consumers. For a level of income M0, utility is
given as U0. The price of c is normalised to unity and bid–rent for the
house is B. It is also conceivable that a range of financial benefits and sav-
ings per annum (S) may accrue to the consumer due to superior energy
performance. Such monetary savings generated by energy efficient fea-
tures will be a function of the present value of the energy cost savings.
However, this is dependent upon uncertain assumptions about future en-
ergy price inflation, behavioural patterns and appropriate discount rates.
Let us assume, therefore, that such expected savings are associated with
a probability distribution. We also include the additional cost (t) of
obtaining energy efficient features and appliances to comply with regula-
tory standards. So the individual's budget constraint can be written as:
n
B þ c þ t–Sð1−∂Þ ¼ M:
The bid–rent, B, is the consumer's willingness to pay for h. ∂ is the dis-
count rate and n is number of years over which savings will accrue.
Therefore, Eq. (1) may be represented by:
U ¼ U fhð:Þ; ðM þ EðSÞ−B−t Þg

n 
¼ U a; f ; r; Lð:Þ; Q ð:Þ; Eð:Þ; M þ Sð1−∂Þ −B t :
Inverting Eq. (7) leads to a bid–rent function as follows,

n 
B ¼ B a; f ; r; Lð:Þ; Q ð:Þ; Eð:Þ; M þ Sð1−∂Þ −t :
Using any attribute of h(.), the shape of the bid–rent function in
Eq. (5) is then represented by the partial derivatives e.g., with respect
to EPC ratings.
BE ¼ U E =U c :
The assumption of a strictly diminishing marginal rate of substitu-
tion means that B will increase with e, but at a decreasing rate. We
can interpret the slope BE as the price that a consumer is willing to
pay for an incremental increase in the EPC rating of a house or, in
other words, as the hedonic price of EPC ratings. Given this interpreta-
tion, Eq. (6) is analogous to the usual utility maximisation condition
equating the price ratio and the marginal rate of substitution. Eq. (8)
can be estimated using regression techniques and our focus is on price
147
ð2Þ effect or coefficient of each parameter i.e., Eq. (9). Before employing
these regression techniques, there are several possible sources of bias
that may impede the ability to estimate Eq. (8) consistently and these
are described below.
First, the EPC rating, as in Eq. (5), may be associated with the pres-
ence (or absence) of a range of energy efficient features in a dwelling.
Key features tend to be associated with the type and nature of lighting,
ð3Þ
insulation, water heating and glazing. This implies presence of correla-
tion between Q and E in Eq. (8). Given the subjective and multi-
faceted nature of the quality variable Q, there may be unobservable
attributes that are correlated with the observed ones. In other words
there may be unobserved heterogeneity leading to a significant omitted
variable bias. Similarly, location, as reflected in Eq. (3), may be a signif-
ð4Þ icant source of unobserved heterogeneity.
Second, given that energy efficient features generate a blend of direct
and indirect and monetary and non-monetary benefits, in terms of a
priori expectations, it is difficult to quantify S. As a baseline we might ex-
pect that a direct price impact of superior energy performance will result
from the monetary savings generated by energy efficient features, and
ð5Þ this will be a function of the present value of the energy cost savings.
However, even this baseline effect is dependent upon uncertain assump-
tions about future energy price inflation and appropriate discount rates.
Therefore, the budget constraint in Eq. (6) may be mis-specified.
Third, making a reasonable assumption that the energy efficient fea-
tures will tend to be newer, it is expected on average that other life cycle
costs associated with maintenance and replacement will be lower. Even
more difficult to quantify is the value of the hedge that energy efficient
features provide against increases in energy costs and changes to energy
regulation. In addition to these direct financial benefits, dwellings with
energy efficient features are likely to obtain other non-energy related
benefits such as, acoustic and security improvements from energy
efficient glazing and improved thermal comfort. These are essentially
quality effects in that the presence of energy efficient features improves
the overall quality of the dwelling. Such a bundle of positive attributes
has the potential to increase marketability of dwellings, leading to
shorter time on market as well as a price impact. This effect may be
more discernible where there are concentrations of eco-consumers
who, in turn, obtain utility from the presence of energy efficient fea-
tures. The presence of such potentially correlated influences is likely
ð6Þ to prevent a precise and exclusive quantification of the price impact of
an eco-label.
Finally, some variables may call for a non-linear specification. For
example, the price of a dwelling might respond to age in two ways;
physical depreciation may reduce price paid but eventually price may
respond positively to age — a ‘vintage’ effect. Moreover, older buildings
tend to be less energy efficient and may involve significant refurbish-
ð7Þ ment costs in order to comply with contemporary building regulations.
Therefore, there may be a high correlation between age and EPC ratings.
Next, we turn to the empirical estimation framework appropriate for
addressing a number of issues raised above.
ð8Þ
4. Estimation strategy
The empirical estimations of the theoretical model in Eq. (8) can be
undertaken in the following form: we derive the following:
XI
ð9Þ P it ¼ β0 þ i¼1
βi X i þ eit ð10Þ
where Pit is the transaction price of a property (specified as the natural
logarithm of price in £ per square metre), Xi is a vector of several explan-
atory locational and physical variables including our variable of interest,
EPC rating; βi is a vector of coefficients to be estimated; and ei is a ran-
dom error and stochastic disturbance term that is expected to take the
form of a normal distribution with a mean of zero and variance of σ2e.
The hedonic weights assigned to each variable are equivalent to their
overall contribution to price (Rosen, 1974). However, hedonic models
148 F. Fuerst et al. / Energy Economics 48 (2015) 145–156

of property sales almost always involve a time dimension as sale sales of the same dwelling. Specifically, the difference in sale prices be-
transactions are collected over a period of months, quarters or years. tween two transaction dates is regressed on a set of dwelling attributes
To capture this inter-temporal variation, we enrich Eq. (10) with a set that includes the EPC rating in the following form:
of binary variables that represent the average effect of each time period
X XT  X XT 
separately in the following form (see Bailey et al., 1963; Gatzlaff and 2 1
P i −P i ¼
J 2
β j X ji þ cD
2
−
J 1
βX þ
1 21
c D þ ei ð12Þ
j¼1 t¼1 t i j¼1 j ji t¼1 t i
Haurin, 1997):
XJ XT
P it ¼ β j X ji þ cD þ eit ð11Þ where the first and second sale periods are denoted by the superscripts
j¼1 t¼1 t t
1 and 2 respectively. Assuming that most house characteristics remain
the same between two sales of the same house, Eq. (12) simplifies to:
where ct is the additional vector of estimated coefficients for each time
period and Dt is a set of variables that takes the value of 1 if a house is XT  
2 1 2 1 21
sold in the period and 0 if it is not sold. P i −P i ¼ c
t¼1 t
Di −Di þ ei : ð13Þ
For the purpose of this study, we specify hedonic models to explain
two dependent variables; price per square metre and price per square Hence, a ‘pure’ repeat-sale model only requires information on
metre change (appreciation/depreciation). To capture the effects of prices and time of transaction. However, as the mix of properties that
EPC rating on these variables, we use a set of binary variables to indicate are sold in each period changes (for example, large detached houses
the EPC band of each dwelling at the relevant transaction date. Band D is might be transacted more often than other types during certain
the ‘hold-out’ category so the coefficients for the higher bands are ex- periods), it is also necessary to control for hedonic characteristics such
pected to be positive. In addition to mitigating the effects of extreme as size, age, and type.
values, the semi-log specification of the hedonic model allows us to In our specification, we use a regional index to capture ‘expected’
interpret the coefficients as average percentage premiums.1 appreciation following the general regional trend as well as the
The standard hedonic regression model uses price per square metre property-specific price components in the following form:
of the dwelling as the dependent variable and a number of property and
local area attributes as independent variables. However, a common P 2t RI2 X J
problem is lack of control for unobserved heterogeneity. There are ¼ t1 þ X þ u j:
j¼1 jt
ð14Þ
P 1t RIt
two sources of unobserved heterogeneity in Eq. (8): locational attri-
butes L and dwelling quality Q. If these unobserved effects are correlated
Thus price changes between two transactions are driven by the
with observed attributes then the estimates will be biased. One way to
regional or local housing market that a property is located in, the time
control for locational attributes is to include local area fixed effects
elapsed between the two sales and a set of observed and unobserved
(specified as dummy variables) in the model under the assumption
property characteristics that cause a house price to deviate from the
that correlated unobservables are time-invariant. In our cross-section
regional trend. The first factor is captured by the regional index ratio
model, we explicitly control for such unobserved effects. With regard
whilst the observed property-specific factors are represented by the
to dwelling quality, several variables are typically incorporated in resi-
dential hedonic price models that describe objectively the attributes of
a dwelling. These include property type, age and number of bedrooms
Table 1
for example. It is not usually possible to obtain a ‘quality’ rating of a Descriptive statistics for key categorical variables (n = 333,095).
dwelling nor a measure of how this quality might have changed over
Variable Categories Frequency % of total
time, perhaps as a result of home improvements. It appears plausible
that the EPC rating is not a proxy for the overall condition or the visual Property type Detached 85,007 25.52%
Semi-detached 106,793 33.22%
appeal of a property as it is possible to obtain a high EPC rating for a
Terraced 115,793 32.06%
property with poor non-energy related maintenance, decoration and Flat 25,637 7.70%
visual appeal (and vice versa). In the absence of detailed information Tenure Freehold 291,778 87.60%
on the nature of improvements, we assume that they are evenly distrib- Leasehold 41,317 12.40%
uted throughout the stock of dwellings. However, it may be suspected Age of dwelling Missing 49,655 14.91%
Before 1900 31,606 9.79%
that unobserved improvements are correlated with the EPC ratings.
1900–1929 41,095 12.34%
Any such correlation would only be problematic if it concerned non- 1930–1949 37,350 11.21%
energy-related features, such as a new kitchen, bathroom and an exten- 1950–1966 36,488 10.95%
sion to the existing property. By contrast, upgrades such as better doors, 1967–1975 30,874 9.27%
windows and boilers have clear implications for the energy efficiency of 1976–1982 17,557 5.27%
1983–1990 26,518 8.14%
a building even if they may, at the same time, be upgrades in terms of 1991–1995 14,689 4.41%
aesthetics, safety, etc. Hence, even if we could identify improved proper- 1996–2002 27,329 8.20%
ties, it would be difficult to disentangle the energy-related value en- 2003–2006 14,826 4.45%
hancements from those of other dimensions. To ensure the robustness 2007 onwards 3894 1.17%
Number of bedrooms 0 8804 2.64%
of our empirical estimates, we attempt to mitigate the effects of poten-
1 11,839 3.55%
tial omitted variable bias by removing observations that are more likely 2 95,260 28.60%
to have been improved. 3 145,546 43.70%
To measure the influence of EPC rating on price appreciation, we also 4 56,891 17.08%
undertake a hedonic analysis with repeat sales transactions only. Effec- 5 12,164 3.65%
5+ 2951 0.78%
tively, we are controlling for the unit fixed effects by comparing two Energy efficiency band A 8 0.00%
B 5119 1.54%
1 C 79,722 23.93%
The ‘raw’ coefficients of the EPC dummy variables require adjustment to determine
D 150,779 45.27%
the percentage premiums (or discounts) as suggested by Halvorsen and Palmquist
E 77,502 23.27%
(1980) and Giles (2011): our simplified adjustment formula follows the van Garderen
F 16,886 5.07%
and Shah (2002) method where the proportional impact pj of a binary variable on the de-
G 3072 0.92%
pendent variable in a semi-logarithmic regression is computed as: pj = [exp(cj) − 1] with
Missing 7 0.00%
cj being the estimated coefficient of the dummy variable.
 F. Fuerst et al. / Energy Economics 48 (2015) 145–156 149

vector of characteristics X. Finally, unobserved characteristics are Table 3

captured in the error term u. Using this framework we are able to Cross tabulation between dwelling type and EPC rating.

observe and estimate the magnitude price differentials that result EPC rating Detached Semi-detached Terraced Flat
from dwellings being placed in different EPC bands. More specifically, A 1 6 0 1
the empirical modelling will comprise: B 398 645 1422 2654
C 17,677 22,855 28,842 10,348
• Cross-sectional models for price levels disaggregated by property D 41,725 49,248 51,121 8685
type, region and EPC rating. E 18,508 28,307 27,694 2993
F 5810 4972 5293 811
• Repeat sale models for price appreciation disaggregated by property G 884 758 1285 145
type and EPC rating. Missing 4 2 1 0
• Robustness checks for restricted samples. All 85,007 106,793 115,658 25,637

(urban–rural indicator) and Lower Level Super Output Areas. A full list
5. Data and descriptive statistics
of these variables is available in Appendix 1. In the third step, EPC
data, maintained by Landmark on behalf of the Department of Commu-
The economic analysis of house price determination, and the pricing
nities and Local Government, were added to the database. This was
of housing attributes, benefits from a large sample of properties to
carried out using address-matching software. Landmark was unable to
ensure that the findings are representative of the entire market. This is
match transactions to EPC records for Wales so, following the removal
particularly relevant in cases where the variables of interest are expect-
of duplicate records and those that contained input and administrative
ed to have only a moderate impact on prices. This might be because the
errors, the resultant sample included 333,095 dwellings.
relatively weak pricing signal is overwhelmed by idiosyncratic price
Descriptive statistics are provided in Tables 1 and 2 and there are a
components in a small sample with large residual errors or due to unob-
number of notable points. Terraced and semi-detached dwellings ac-
served attributes of a particular property or set of properties. Such
count for approximately one-third of the sample. Detached dwellings
concerns are addressed, at least in part, by obtaining a large sample of
represent around a quarter with flats accounting for just less than 8%
dwelling transactions and associated control variables. In the context
of the total. Most dwellings in the sample are held on freehold tenure.
of this study, the hedonic estimation was dependent upon the availabil-
There is no information on age for nearly 15% of the observations but
ity of data in three main areas; (1) market prices, (2) energy perfor-
where this has been recorded the maturity of the housing stock is
mance and (3) building and location attributes. The collection and
clear; 44% of the sample was built more than 50 years ago. Two variables
assembly of data from these three areas is described below.
measure dwelling size: number of bedrooms, the vast majority (89%) of
Datasets from several sources were merged into a unified database.
dwellings comprising two to four bedrooms with 44% of the sample in
In the first step, data on market prices were obtained from Calnea
the three-bedroom category; and floor area, with a median size of
Analytics, comprising residential transaction prices as submitted to the
85 m2. In terms of energy efficiency, most dwellings (93%) are in
UK Land Registry. To enable repeat-sales as well as pooled cross-
bands C, D or E. Only seven are in band A, which is why we formed a
sectional analysis, the sample contains the prices of dwellings that
combined A/B category for the purpose of this analysis. Table 3 reveals
were sold at least twice in the period 1995–2012. The start of the
that flats tend to be the most energy efficient dwelling type with 50%
study period is determined by the availability of comprehensive attri-
in bands B and C. In contrast, 21% of detached properties were in
bute data. The second transaction in each pair of sales is determined
bands B or C.
by the availability of energy labelling information to ensure that an
The impact of age on EPC band is shown in Fig. 1; more modern
EPC rating was available at the time of at least one transaction for
dwellings, built since 2003, are in band B, those built in the 1990s
each dwelling. This effectively means that at least one of the transac-
tend to be found in bands C and D whereas older dwellings are predom-
tions must have occurred after August 2008 when EPCs became manda-
inantly found in bands E, F and G. Combining EPC band with the size and
tory for all residential transactions in the UK. The sample was further
age variables and the urban/rural indicator variable (which scores each
refined by the availability of essential information on property location,
dwelling according to the extent to which it is located in an urbanised
type and size. This information is captured in the Calnea database
area), reveals that older, larger dwellings located in rural areas tend to
through both estate agency listings and surveyor visits. Applying these
have lower EPC ratings than smaller, modern dwellings in urban areas.
criteria, we obtained a sample of 500,000 dwellings and one million
transaction prices in England and Wales, randomly drawn from a pool
of approximately five million transactions that match the above criteria. Dwellings in each EPC band (B to G) classified by age band
No transaction prices and/or EPC information were available for 100%
Scotland and Northern Ireland. G
90%
In the next step of data assembly, we obtained and matched socio- F
80%
economic data from the Office for National Statistics as well as a series
70% E
of indicators collected and published by the UK Census using a
Geographic Information System. The geographic reference of these 60%
D
area variables is a combination of postcode districts, Output Areas 50%
C
40%

30% B
Table 2
20%
Descriptive statistics for key continuous variables (n = 333,095).
10%
Variable Mean SD
0%
Price (P1) 162,285 133,188
Price (P2) 226,696 191,172
Compound annual growth rate (%) 4.76% 6.40%
Total floor area (m2) 95 45
Energy efficiency rating 60 12
Fig. 1. Dwellings in each EPC band (B to G) classified by age band.
150 F. Fuerst et al. / Energy Economics 48 (2015) 145–156

Table 4
Energy rating and price: hedonic estimations (dependent variable: log of price per square metre).

(1) (2) (3) (4) (5) (6) (7)

Full sample Semi-detached Terraced Flat Detached dense Detached sparse Full sample

EPC A/B 0.050⁎⁎⁎ 0.008 0.045⁎⁎⁎ 0.016⁎ −0.020 −0.111⁎⁎

(9.93) (0.49) (4.67) (2.21) (−0.97) (−2.61) –
EPC C 0.018⁎⁎⁎ 0.005⁎ 0.015⁎⁎⁎ 0.008⁎ −0.002 −0.031⁎⁎⁎ –
(15.49) (2.46) (7.68) (2.14) (−0.78) (−3.87)
EPC D Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out –
EPC E −0.007⁎⁎⁎ −0.013⁎⁎⁎ −0.019⁎⁎⁎ −0.014⁎⁎ −0.012⁎⁎⁎ 0.008 –
(−6.10) (−7.07) (−9.88) (−2.60) (−4.24) (1.22)
EPC F −0.009⁎⁎⁎ −0.023⁎⁎⁎ −0.046⁎⁎⁎ −0.029⁎⁎ −0.020⁎⁎⁎ 0.009 –
(−3.81) (−5.20) (−10.90) (−3.00) (−3.49) (1.03)
EPC G −0.068⁎⁎⁎ −0.065⁎⁎⁎ −0.121⁎⁎⁎ −0.072⁎⁎ −0.070⁎⁎⁎ 0.018 –
(−10.26) (−5.33) (−11.98) (−2.81) (−3.37) (1.06)
Energy efficiency score – – – – – – 0.0006⁎⁎⁎
(20.25)
No. of beds −0.043⁎⁎⁎ −0.041⁎⁎⁎ −0.041⁎⁎⁎ −0.049⁎⁎⁎ −0.035⁎⁎⁎ −0.030⁎⁎⁎ −0.043⁎⁎⁎
(−71.26) (−37.48) (−37.19) (−17.47) (−31.80) (−14.20) (−71.20)
Pre-1900 Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out
1900–1929 −0.085⁎⁎⁎ −0.075⁎⁎⁎ −0.068⁎⁎⁎ −0.050⁎⁎⁎ −0.029⁎⁎ −0.065⁎⁎⁎ −0.085⁎⁎⁎
(−39.35) (−15.48) (−27.07) (−4.38) (−3.19) (−4.33) (−39.70)
1930–1949 −0.051⁎⁎⁎ −0.076⁎⁎⁎ −0.034⁎⁎⁎ −0.052⁎⁎⁎ −0.007 −0.066⁎⁎⁎ −0.053⁎⁎⁎
(−22.88) (−17.10) (−10.28) (−4.22) (−0.82) (−4.88) (−23.57)
1950–1966 −0.055⁎⁎⁎ −0.080⁎⁎⁎ −0.090⁎⁎⁎ −0.135⁎⁎⁎ 0.024⁎ −0.048⁎⁎⁎ −0.058⁎⁎⁎
(−24.22) (−17.53) (−25.85) (−12.25) (2.44) (−4.51) (−25.17)
1967–1975 −0.052⁎⁎⁎ −0.041⁎⁎⁎ −0.095⁎⁎⁎ −0.132⁎⁎⁎ −0.034⁎⁎⁎ −0.110⁎⁎⁎ −0.054⁎⁎⁎
(−22.25) (−8.55) (−27.32) (−12.03) (−4.03) (−10.55) (−23.24)
1976–1982 −0.023⁎⁎⁎ −0.027⁎⁎⁎ −0.013⁎⁎ −0.060⁎⁎⁎ −0.049⁎⁎⁎ −0.125⁎⁎⁎ −0.026⁎⁎⁎
(−8.86) (−5.19) (−3.13) (−5.19) (−5.75) (−9.86) (−9.87)
1983–1990 0.029⁎⁎⁎ 0.022⁎⁎⁎ 0.080⁎⁎⁎ −0.0001 −0.04⁎⁎⁎ −0.100⁎⁎⁎ 0.026⁎⁎⁎
(11.94) (4.41) (22.53) (−0.01) (−4.78) (−8.39) (10.71)
1991–1995 0.044⁎⁎⁎ 0.035⁎⁎⁎ 0.115⁎⁎⁎ 0.020 −0.036⁎⁎⁎ −0.070⁎⁎⁎ 0.041⁎⁎⁎
(16.20) (6.37) (26.31) (1.65) (−4.25) (−5.24) (14.93)
1996–2002 0.026⁎⁎⁎ 0.035⁎⁎⁎ 0.093⁎⁎⁎ 0.039⁎⁎⁎ −0.031⁎⁎⁎ −0.100⁎⁎⁎ 0.024⁎⁎⁎
(10.71) (6.68) (22.71) (3.33) (−3.78) (−9.39) (9.66)
2003–2006 0.018⁎⁎⁎ 0.008 0.061⁎⁎⁎ 0.032⁎⁎ −0.033⁎⁎⁎ −0.101⁎⁎⁎ 0.020⁎⁎⁎
(6.05) (1.21) (11.75) (2.63) (−3.74) (−7.95) (6.56)
2007– 0.022⁎⁎⁎ −0.0002 0.082⁎⁎⁎ 0.015 0.004 −0.075⁎⁎⁎ 0.029⁎⁎⁎
(3.78) (−0.02) (6.75) (0.90) (0.36) (−3.43) (5.06)
Age (unknown) −0.055⁎⁎⁎ −0.074⁎⁎⁎ −0.036⁎⁎⁎ −0.0488⁎⁎⁎ −0.0640⁎⁎⁎ −0.138⁎⁎⁎ −0.056⁎⁎⁎
(−24.73) (−15.62) (−11.75) (−4.33) (−7.77) (−13.86) (−25.02)
Freehold 0.063⁎⁎⁎ 0.038⁎⁎⁎ 0.081⁎⁎⁎ 0.065⁎⁎⁎ 0.022⁎⁎ 0.090⁎ 0.065⁎⁎⁎
(21.34) (7.74) (16.71) (5.92) (3.22) (2.58) (21.23)
Deprivation score −0.009⁎⁎⁎ −0.009⁎⁎⁎ −0.009⁎⁎⁎ −0.007⁎⁎⁎ −0.005⁎⁎⁎ −0.002⁎ −0.009⁎⁎⁎
(−161.27) (−99.20) (−97.33) (−33.84) (−33.18) (−2.18) (−161.14)
Urban level Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out
Category 1
Urban level −0.013 0.011 −0.086 0.187 0.063 . −0.013
Category 2 (−0.58) (0.24) (−1.62) (1.60) (0.98) . (−0.59)
Urban level 0.079⁎⁎⁎ 0.043 0.084 0.259⁎ . −0.004 0.081⁎⁎⁎
Category 3 (3.74) (1.00) (1.68) (2.29) . (−0.19) (3.82)
Urban level 0.092⁎⁎⁎ 0.054 −0.031 0.209 . . 0.097⁎⁎⁎
Category 4 (3.74) (0.87) (−0.50) (1.55) . . (3.84)
Urban level −0.052⁎ −0.019 −0.091 0.096 0.024 . −0.052⁎
Category 5 (−2.38) (−0.42) (−1.77) (0.90) (0.40) . (−2.37)
Urban level −0.040 −0.012 −0.061 0.072 0.044 . −0.040
Category 6 (−1.82) (−0.25) (−1.34) (0.67) (0.72) . (−1.81)
Urban level 0.032 0.041 −0.003 0.091 . −0.058⁎ 0.033
Category 7 (1.46) (0.90) (−0.06) (0.84) . (−2.42) (1.52)
Urban level 0.058⁎⁎ 0.062 −0.013 0.076 . −0.027 0.060⁎⁎
Category 8 (2.64) (1.33) (−0.25) (0.69) . (−1.12) (2.71)
Terraced Hold-out . . . Hold-out
Detached 0.186⁎⁎⁎ . . . 0.186⁎⁎⁎
(124.97) (125.43)
Semi-detached 0.089⁎⁎⁎ 0.089⁎⁎⁎
(75.11) (75.31)
Flat −0.002 0.0001
(−0.61) (0.03)
Constant 7.944⁎⁎⁎ 7.985⁎⁎⁎ 7.992⁎⁎⁎ 8.092⁎⁎⁎ 7.951⁎⁎⁎ 7.890⁎⁎⁎
(356.84) (170.98) (153.41) (75.83) (126.67) (352.23)

Quarterly fixed effects Yes Yes Yes Yes Yes Yes Yes
Postcode fixed effects Yes Yes Yes Yes Yes Yes Yes
Adj. R2 0.693 0.651 0.785 0.721 0.587 0.470 0.693
N 333,095 106,793 115,658 25,637 68,354 16,653 333,095
 F. Fuerst et al. / Energy Economics 48 (2015) 145–156
This finding from the descriptive statistics is borne out in the hedonic
modelling below.
6. Findings
Following the analytical strategy and data sampling outlined above,
we first fit regression models to both the full set of observations and the
sub-samples of the different types of dwelling. The results are presented
in Table 4. The log of dwelling price per square metre is explained as a
function of four dwelling attributes (age, dwelling type, number of
bedrooms and tenure), two composite neighbourhood attributes
(urban–rural index score and deprivation index score), quarterly time
fixed effects, postcode area fixed effects and EPC rating. In terms of
explanatory power, the adjusted R2 is nearly 70% for the full sample
and the coefficients of the independent variables largely have the
expected signs. The ‘number of bedrooms’ coefficient is negative and
highly significant. By expressing the dependent variable as price per
square metre rather than total price, we largely eliminate the size effect
of a property from the estimation. Whilst this reduces the predictive
power of the model and may not represent how the majority of buyers
in the UK think about property prices, it presents a purer measure of
prices. The bedroom coefficient then captures both the negative effect
of smaller room sizes and possibly also diminishing marginal utility
from consuming more space, i.e., all else equal, larger properties with
more bedrooms command lower prices per square metre than smaller
properties. The effect of age on dwelling price per square metre is
non-linear and variable between dwelling types. Compared to dwell-
ings constructed pre-1900, dwellings constructed since 1983 have
sold for small but statistically significant price premiums. When we
look at the results across dwelling types, it is apparent that there are
notable differences between semi-detached, terraced and detached
properties. In contrast to semi-detached and terraced dwellings, de-
tached properties constructed in the last three decades tend to sell for
significantly less per square metre than dwellings constructed before
1900. The results for dwelling type are also in line with expectations.
With terraced dwellings as the ‘hold-out’ category, semi-detached and
detached properties achieve significantly higher prices per square
metre, with the latter selling for approximately 19% more per square
metre than terraced dwellings. The coefficients for deprivation and
rural indexes also have the expected signs. Compared to leasehold, the
coefficient for freehold is positive and significant.
Turning to the variable of interest, EPC rating, and using band D as
the ‘hold-out’ category, the pattern of price effects reveals a positive
relationship between energy performance rating and sale price. For
the whole sample model, there are significant positive premiums for
dwellings rated A/B (5%) or C (1.8%) compared to dwellings rated D.
For dwellings rated E (−0.7%) and F (−0.9%), there are statistically sig-
nificant discounts. The relative price effects are highest for terraced
dwellings. All else equal, we estimate that a terraced dwelling rated B
has sold for approximately 4.5% more per square metre than a terraced
dwelling EPC rated D (see Column 4 in Table 4). The comparable figure
for a flat is 1.6%. Compared to D rated dwellings, all dwelling types rated
E, F and G sold at statistically significant discounts. In Table 4 (Column
7) the results of the estimation when energy efficiency score, rather
than band, is used as the independent variable and the expected posi-
tive relationship between energy efficiency and dwelling sale price is
also found. A 1% increase in the energy efficiency score produces a
0.1% increase in predicted dwelling price. This price effect can be com-
pared to two previous studies that examined the price effect of energy
efficiency measured on a continuous scale. First Cajias and Piazolo
Notes to Table 4:
t statistics based on robust standard errors are reported within in parentheses.
⁎ p b 0.05.
⁎⁎ p b 0.01.
⁎⁎⁎ p b 0.001.
151
(2013) found that a 1% increase in energy efficiency produced a 0.45%
increase in market value and, second, Hyland et al. (2013) reported a
price fall of 1.3% for each 1-point decline in building energy rating
as measured on a 15-point scale (i.e., a 1% fall in the energy rating
produced a 0.2% fall in price).
With the exception of detached houses, a pattern of increasing price
premiums with increasing energy performance is found for all the
dwelling types. For detached dwellings, no significant price effects
were observed. This apparent anomaly seems to be driven by a relative-
ly small section of the sample consisting of just over 16,653 dwellings in
rural areas. When the detached dwellings are separated into dwellings
located in sparsely populated areas and dwellings located in densely
populated areas, we find that the pattern of price discounts found in
the rest of the sample is replicated for the detached dwellings in densely
populated areas. More specifically, the pattern of price effects for the
68,354 detached dwellings in densely populated areas is very similar
to the pattern of price discounts for the 106,793 semi-detached dwell-
ings. It is also notable that the explanatory power of the hedonic
model is lowest (R2 = 47%) for the sub-sample of 16,653 detached
dwellings in sparsely populated areas. This is likely to be due to the
greater heterogeneity of this particular group, which will include large
country residences together with a wide range of rural dwellings, built
in vernacular styles over several centuries.
We also apply a similar regression specification with dwelling price
appreciation per square metre as the dependent variable. We do not
have definite prior expectations for either positive or negative effects.
It is possible that price premiums associated with superior energy per-
formance have been factored into initial prices and that there is no
‘growth premium’. On the other hand, it is possible that the increasing
salience of energy and environmental issues in the last decade has
meant that price effects have produced positive effects on price appreci-
ation. In other words, the effects of superior energy performance on ini-
tial prices may be positive and, due to subsequent greater demand for
energy efficient dwellings, the effects on price appreciation may also
be positive.
Table 5 provides estimates of the determinants of the dwelling price
appreciation. We see that, for all types of dwelling, number of bedrooms
has a significant positive effect on growth rate. Compared to dwellings
built pre-1900, the prices of dwellings constructed between 1967 and
2007 have appreciated at a significantly lower rate. In contrast, dwell-
ings constructed prior to 1949 have experienced slightly – but statisti-
cally significant – higher appreciation rates compared to the ‘hold-out’
category (dwellings constructed pre-1900), albeit the coefficients are
not significant when the dwellings are disaggregated into types. Given
the sample period and the over-supply of apartments in many markets,
it is perhaps not surprising that, compared to terraced dwellings, flats
have experienced significantly lower rates of price appreciation. It is
likely that housing market sub-sectors may react differently to cyclical
fluctuations, e.g., flats may experience more volatility compared to
other sectors (see Costello, 2000 for example). Our analysis at the sub-
sector level is suitable to alleviate such concerns as we also explicitly
control for the time-specific effects within each sector. Overall, on a
per square metre basis, flats tend to sell for less than other dwelling
types and have experienced lower growth rates. Similarly, freehold
dwellings have sold for higher prices per square metre compared to
leasehold dwellings and have experienced a significantly higher rate
of price appreciation.
Turning to the variable of interest, the results for the price apprecia-
tion per square metre model are not as consistent as the price model.
For the full sample, C-rated dwellings have experienced significantly
152 F. Fuerst et al. / Energy Economics 48 (2015) 145–156

higher price appreciation than D-rated dwellings. However, this is not bands E and F are also estimated to have appreciated at significantly
the case for the dwellings in the A/B band, which have experienced sig- lower rates than properties in band D. When we look at the estimates
nificant price depreciation compared to D-rated dwellings. Dwellings in for the dwelling type sub-samples, we see that the effects on price

Table 5
Energy rating and price appreciation: repeat sales estimations (dependent variable: change in price per square metre).
(1) (2) (3) (4) (5) (6) (7)

Full sample Semi-detached Terraced Flat Detached dense Detached sparse Full sample

EPC A/B −0.015⁎⁎⁎ −0.023⁎⁎ −0.003 0.0004 0.042⁎ 0.010 –

(−4.45) (−2.71) (−0.55) (0.06) (2.25) (0.20)
EPC C 0.013⁎⁎⁎ 0.015⁎⁎⁎ 0.005⁎⁎⁎ −0.0001 0.023⁎⁎⁎ 0.019⁎ –
(13.34) (8.57) (3.34) (−0.04) (10.18) (2.49)
EPC D Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out –
EPC E −0.018⁎⁎⁎ −0.025⁎⁎⁎ −0.014⁎⁎⁎ −0.0008 −0.029⁎⁎⁎ −0.032⁎⁎⁎ –
(−18.52) (−15.64) (−8.39) (−0.18) (−12.88) (−5.72)
EPC F −0.026⁎⁎⁎ −0.032⁎⁎⁎ −0.022⁎⁎⁎ 0.005 −0.037⁎⁎⁎ −0.060⁎⁎⁎ –
(−13.76) (−9.39) (−6.60) (0.72) (−8.11) (−8.82)
EPC G −0.001 0.001 0.009 0.013 −0.017 −0.059⁎⁎⁎ –
(−0.21) (0.10) (1.05) (0.61) (−1.03) (−5.02)
Energy efficiency score – – – – – – 0.0014⁎⁎⁎
(22.57)
Regional price 2.373⁎⁎⁎ 2.418⁎⁎⁎ 2.483⁎⁎⁎ 2.293⁎⁎⁎ 2.192⁎⁎⁎ 2.270⁎⁎⁎ 2.373⁎⁎⁎
Index (979.37) (604.66) (582.22) (236.21) (436.40) (168.63) (979.37)
No. of bedrooms 0.011⁎⁎⁎ 0.015⁎⁎⁎ 0.011⁎⁎⁎ 0.009⁎⁎⁎ 0.009⁎⁎⁎ 0.003 0.011⁎⁎⁎
(23.93) (17.57) (12.85) (4.00) (9.68) (1.72) (23.93)
Pre-1900 Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out
1900–1929 0.006⁎⁎⁎ 0.006 0.0001 0.013 0.006 −0.008 0.006⁎⁎⁎
(3.40) (1.68) (0.06) (1.39) (0.82) (−0.72) (3.40)
1930–1949 0.004⁎ −0.002 −0.003 0.013 −0.008 0.016 0.004⁎
(1.97) (−0.70) (−1.11) (1.35) (−1.21) (1.38) (1.97)
1950–1966 −0.001 −0.015⁎⁎⁎ 0.001 0.019⁎ −0.019⁎⁎ 0.012 −0.001
(−0.64) (−4.18) (0.31) (2.24) (−2.95) (1.42) (−0.64)
1967–1975 −0.035⁎⁎⁎ −0.050⁎⁎⁎ −0.030⁎⁎⁎ −0.007 −0.053⁎⁎⁎ −0.037⁎⁎⁎ −0.035⁎⁎⁎
(−19.75) (−14.06) (−10.97) (−0.86) (−8.52) (−4.54) (−19.75)
1976–1982 −0.053⁎⁎⁎ −0.062⁎⁎⁎ −0.043⁎⁎⁎ −0.028⁎⁎ −0.082⁎⁎⁎ −0.061⁎⁎⁎ −0.053⁎⁎⁎
(−27.45) (−16.05) (−14.36) (−3.21) (−12.69) (−6.17) (−27.45)
1983–1990 −0.071⁎⁎⁎ −0.076⁎⁎⁎ −0.057⁎⁎⁎ −0.050⁎⁎⁎ −0.097⁎⁎⁎ −0.099⁎⁎⁎ −0.071⁎⁎⁎
(−40.37) (−20.60) (−22.17) (−6.20) (−15.38) (−11.35) (−40.37)
1991–1995 −0.089⁎⁎⁎ −0.090⁎⁎⁎ −0.070⁎⁎⁎ −0.074⁎⁎⁎ −0.111⁎⁎⁎ −0.101⁎⁎⁎ −0.089⁎⁎⁎
(−43.97) (−21.72) (−22.67) (−8.48) (−17.04) (−9.51) (−43.97)
1996–2002 −0.124⁎⁎⁎ −0.122⁎⁎⁎ −0.109⁎⁎⁎ −0.120⁎⁎⁎ −0.141⁎⁎⁎ −0.104⁎⁎⁎ −0.124⁎⁎⁎
(−63.89) (−30.06) (−35.04) (−13.89) (−22.43) (−11.15) (−63.89)
2003–2007 −0.116⁎⁎⁎ −0.133⁎⁎⁎ −0.115⁎⁎⁎ −0.117⁎⁎⁎ −0.135⁎⁎⁎ −0.080⁎⁎⁎ −0.116⁎⁎⁎
(−49.75) (−29.41) (−32.38) (−12.86) (−19.28) (−6.66) (−49.75)
2007–2011 −0.048⁎⁎⁎ −0.079⁎⁎⁎ −0.068⁎⁎⁎ −0.076⁎⁎⁎ −0.044⁎⁎⁎ 0.014 −0.048⁎⁎⁎
(−10.69) (−9.22) (−9.97) (−6.77) (−3.83) (0.61) (−10.69)
Age (unknown) −0.034⁎⁎⁎ −0.030⁎⁎⁎ −0.024⁎⁎⁎ −0.032⁎⁎⁎ −0.070⁎⁎⁎ −0.028⁎⁎⁎ −0.034⁎⁎⁎
(−19.28) (−8.41) (−9.91) (−3.78) (−11.08) (−3.47) (−19.28)
Terraced Hold-out Hold-out
Detached −0.012⁎⁎⁎ −0.012⁎⁎⁎
(−10.29) (−10.08)
Semi-detached 0.003⁎⁎⁎ 0.003⁎⁎⁎
(3.57) (3.52)
Flat −0.036⁎⁎⁎ −0.038⁎⁎⁎
(−14.85) (−15.84)
Freehold 0.017⁎⁎⁎ 0.013⁎⁎⁎ 0.018⁎⁎⁎ 0.057⁎⁎⁎ 0.023⁎⁎⁎ 0.050⁎ 0.017⁎⁎⁎
(8.30) (3.84) (5.39) (6.62) (3.89) (2.10) (8.36)
Deprivation 0.0005⁎⁎⁎ 0.0005⁎⁎⁎ 0.001⁎⁎⁎ 0.0002 −0.001⁎⁎⁎ 0.001 0.0006⁎⁎⁎
Index (11.52) (6.19) (8.75) (1.47) (−5.73) (1.23) (11.80)
Urban level Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out
Category 1
Urban level 0.009 0.031 0.015 −0.110 −0.0303 . 0.010
Category 2 (0.56) (1.06) (0.41) (−1.54) (−0.72) . (0.65)
Urban level 0.056⁎⁎⁎ 0.098⁎⁎⁎ 0.075⁎ −0.078 . −0.014 0.059⁎⁎⁎
Category 3 (3.73) (3.44) (2.20) (−0.98) . (−0.81) (3.93)
Urban level 0.056⁎⁎ 0.07 0.097⁎ −0.145 . . 0.060⁎⁎⁎
Category 4 (3.10) (1.66) (2.21) (−1.20) . . (3.31)
Urban level −0.007 0.039 0.042 −0.038 −0.074 . −0.005
Category 5 (−0.44) (1.26) (1.19) (−0.73) (−1.89) . (−0.35)
Urban level −0.001 0.043 0.050 −0.012 −0.068 . −0.00004
Category 6 (−0.09) (1.48) (1.47) (−0.23) (−1.75) . (−0.00)
Urban level 0.015 0.064⁎ 0.047 −0.033 . −0.045⁎ 0.017
Category 7 (0.96) (2.07) (1.33) (−0.62) . (−2.33) (1.09)
Urban level 0.022 0.074⁎ 0.047 −0.030 . −0.029 0.024
Category 8 (1.39) (2.36) (1.30) (−0.55) . (−1.47) (1.56)
Intercept 0.035⁎ −0.024 −0.033 0.076 0.137⁎⁎⁎ 0.079⁎ −0.020
(2.24) (−0.77) (−0.92) (1.45) (3.39) (2.13) (−1.24)
 F. Fuerst et al. / Energy Economics 48 (2015) 145–156
Table 5 (continued)
(1) (2)
Full sample Semi-detached
Quarterly fixed effects Yes Yes
Adj. R2 0.788 0.807
N 333,004 106,772
t statistics based on robust standard errors are reported within parentheses.
⁎ p b 0.05.
⁎⁎ p b 0.01.
⁎⁎⁎ p b 0.001.
appreciation are largest for detached dwellings and that there is no
significant effect on flats rated E and F. In Table 5, Column 7 displays
the results of the estimation when energy efficiency score, rating than
band, is used as the independent variable. The expected positive rela-
tionship between energy efficiency and price appreciation is also found.
7. Robustness checks
As noted above, a common issue with hedonic estimations is poten-
tial omitted variable bias. In the context of this research one particular
concern is that dwellings with better EPC ratings may have been subject
to unobserved improvements that enhance the quality of dwellings in
addition to enhancing energy performance. In order to try to counteract
such potential bias, we run the models with a number of restricted sam-
ples. The main purpose of the restrictions is to exclude dwellings that
are more likely to have been improved or that may be unusual in
some way; dwellings that have been re-sold within a short period of
time or dwellings exhibiting high levels of unexplained variance in the
estimations for example. Firstly, we restrict the sample to dwellings
sold where the EPC is known at time of sale. Second, we remove obser-
vations that had a relatively high level of unexplained variation in the
full sample model. The rationale behind this test is that the dwellings
that might have been improved are more likely to have a high level of
unexplained price variation in our regression specification. Therefore,
once we have run the full sample model, we then identified properties
with a large discrepancy between the price predicted by the model
and the observed price, and then re-estimate the model without those
properties in the regression sample. This two-stage estimation is analo-
gous to estimation techniques used in spatial econometrics to identify
unobserved spatial effects (see for example Case et al., 2004), and
should provide us with results from a sample of properties whose
price variations are reasonably explained by the observable informa-
tion. Third, a related concern is the possibility of sample selection bias
in our sample of transacted properties. A well-established literature ar-
gues that more frequently traded residential properties are of lower
quality (Bourassa et al., 2006; Chernobai and Chernobai, 2013;
Costello and Watkins, 2002; Englund et al., 1999; Jansen et al., 2008).
In our context, it is likely that frequently transacted properties may
have undergone energy efficiency-related improvements. However,
our sample includes dwellings that have been sold twice in the time pe-
riod under investigation, namely 17 years. According to the English
Housing Survey (2014) the average holding period for an owner-
occupied dwelling in 17.3 years. Given the long time period for inclusion
in our sample, it is unlikely that the results will reflect any significant
sample selection bias. However, in order to address such concern, we
remove observations where the property had been sold again within
two years.
It is reassuring that the results remain broadly stable. The details are
presented in Table 6. For the cross-sectional price models the results of
the three restricted sample models estimate similar patterns of pre-
miums and discounts compared to EPC D as the full sample model
with one exception; A/B rated dwellings in the sample restricted to
dwellings where the EPC is known at time of both sales. It should be
153
(3) (4) (5) (6) (7)
Terraced Flat Detached dense Detached sparse Full sample
Yes Yes Yes Yes Yes
0.799 0.789 0.779 0.707 0.788
115,604 25,635 68,346 16,647 333,004
noted that this restriction resulted in a small sample size of 18,135
dwellings. Whilst restricting the sample to dwellings with low unex-
plained variance reduces the sample by approximately 80%, the
estimates of price premiums and discounts remain indistinguishable
from the full sample results. Although the outcome of this robustness
test is supportive of the main findings, the caveat of omitted variable
bias can still be significant to consider when interpreting the parameter
estimates. Similarly, taking out dwellings that have been ‘flipped’ in
under two years has no notable effect on the estimated coefficients.
Applying similar restrictions to the data applied to price appreciation
results in a similar pattern with very little variation in the estimated
coefficients for the various restricted samples.
It was estimated above that the largest relative price effects associat-
ed with EPC rating were found with terraced dwellings. This may be due
to the fact that terraced properties tend to be cheaper than other dwell-
ing types. This, in turn, generates an expectation that the absolute value
of energy savings associated with enhanced energy performance will
have different relative price effects. Given marked regional variations
in house prices, assuming that price premiums or discounts to some
extent represent the present value of expected energy savings, there
should also be a negative relationship between house prices and
percentage price premium or discount. Table 7 presents the results of
the hedonic estimations disaggregated by region.2 The pattern of price
effects is broadly as expected. The largest premiums and discounts relat-
ed to dwellings rated D are in the region with the lowest house prices —
the North East. In regions with the highest house prices (London, South
East, South West and East Anglia) there are either lower or no statisti-
cally significant price premiums/discounts. Thus, the findings from a re-
gional disaggregation are consistent with the expectation that similar
absolute values of energy savings and other benefits associated with
good energy ratings will have different relative effects on house prices.
8. Conclusion
Reflecting growing concern about man-made climate change, over
the past decade there has been an increasing policy focus on improving
the environmental performance of the housing stock. The core function
of mandatory energy efficiency certification in the EU has been to
change consumer behaviour by providing reliable information on the
energy performance of dwellings to buyers. Given the quasi-credence
attributes of residential property, it is often not feasible for consumers
to directly measure some desired characteristics such as energy efficien-
cy. The key assumption underpinning the introduction of market-based
policy instruments such as certification is that energy efficient attributes
will be capitalised which will, in turn, stimulate increased supply of new
energy efficient dwellings and the refurbishment of existing dwellings
to improve their energy performance.
Drawing upon a large sample of dwellings with mandatory energy
certificates, we find that the vast majority of dwellings are clustered in
the middle EPC bands (C, D and E). Nearly half of all dwellings are
2
There are nine regions in England; London, South East, South West, East Midland,
West Midlands, East of England, Yorkshire and Humberside, North East and North West.
154 F. Fuerst et al. / Energy Economics 48 (2015) 145–156

Table 6
Energy rating and price: restricted sample results.

Price PSM Price appreciation PSM

EPC known at both transactions Low residual Re-sale N 2 years EPC known at both transactions Low residual Re-sale N 2 years

EPC A/B 0.098⁎⁎⁎ 0.052⁎⁎⁎ 0.041⁎⁎⁎ −0.026⁎⁎⁎ −0.015⁎⁎⁎ −0.015⁎⁎⁎

(6.65) (46.49) (7.40) (−4.28) (−18.31) (−3.55)
EPC C 0.020⁎⁎⁎ 0.019⁎⁎⁎ 0.019⁎⁎⁎ 0.006 0.012⁎⁎⁎ 0.014⁎⁎⁎
(3.29) (57.76) (15.17) (1.70) (47.76) (13.31)
EPC D Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out
EPC E −0.010 −0.007⁎⁎⁎ −0.007⁎⁎⁎ −0.006⁎ −0.017⁎⁎⁎ −0.019⁎⁎⁎
(−1.65) (−21.12) (−5.74) (−1.98) (−67.74) (−18.47)
EPC F −0.014 −0.009⁎⁎⁎ −0.009⁎⁎ 0.023⁎⁎⁎ −0.025⁎⁎⁎ −0.032⁎⁎⁎
(−1.45) (−14.50) (−3.28) (4.43) (−51.53) (−15.36)
EPC G −0.058⁎⁎⁎ −0.067⁎⁎⁎ −0.059⁎⁎⁎ 0.086⁎⁎⁎ 0.0002 −0.017⁎⁎
(−3.51) (−45.91) (−7.84) (10.70) (0.15) (−2.81)
1.790⁎⁎⁎ 2.371⁎⁎⁎ 2.384⁎⁎⁎
Regional price index – – – (21.45) (3737.76) (949.31)
No. of beds −0.058⁎⁎⁎ −0.043⁎⁎⁎ −0.042⁎⁎⁎ −0.001 0.011⁎⁎⁎ 0.012⁎⁎⁎
(−18.66) (−291.79) (−67.53) (−0.67) (93.14) (24.18)
Pre-1900 Hold-out Hold-out Hold-out
1900–1929 −0.088⁎⁎⁎ −0.083⁎⁎⁎ −0.085⁎⁎⁎ 0.014⁎ 0.005⁎⁎⁎ 0.005⁎
(−8.10) (−151.86) (−38.12) (2.32) (10.75) (2.56)
1930–1949 −0.066⁎⁎⁎ −0.049⁎⁎⁎ −0.051⁎⁎⁎ 0.010 0.003⁎⁎⁎ 0.002
(−6.05) (−85.40) (−22.20) (1.80) (6.76) (0.99)
1950–1966 −0.046⁎⁎⁎ −0.054⁎⁎⁎ −0.057⁎⁎⁎ 0.002 −0.002⁎⁎⁎ −0.003
(−4.27) (−91.39) (−23.89) (0.37) (−3.94) (−1.44)
1967–1975 −0.043⁎⁎⁎ −0.051⁎⁎⁎ −0.054⁎⁎⁎ −0.013⁎ −0.035⁎⁎⁎ −0.038⁎⁎⁎
(−3.82) (−85.59) (−22.18) (−2.22) (−74.42) (−20.17)
1976–1982 0.008 −0.020⁎⁎⁎ −0.025⁎⁎⁎ −0.041⁎⁎⁎ −0.052⁎⁎⁎ −0.055⁎⁎⁎
(0.60) (−29.56) (−9.35) (−6.17) (−98.02) (−27.01)
1983–1990 0.032⁎⁎ 0.031⁎⁎⁎ 0.027⁎⁎⁎ −0.057⁎⁎⁎ −0.069⁎⁎⁎ −0.073⁎⁎⁎
(2.63) (48.64) (10.96) (−9.15) (−142.88) (−39.37)
1991–1995 0.063⁎⁎⁎ 0.045⁎⁎⁎ 0.042⁎⁎⁎ −0.075⁎⁎⁎ −0.086⁎⁎⁎ −0.091⁎⁎⁎
(4.29) (62.16) (15.24) (−10.35) (−156.15) (−42.63)
1996–2002 0.038⁎⁎ 0.028⁎⁎⁎ 0.025⁎⁎⁎ −0.081⁎⁎⁎ −0.119⁎⁎⁎ −0.126⁎⁎⁎
(2.78) (43.53) (10.00) (−11.54) (−235.25) (−62.16)
2003–2006 0.016 0.019⁎⁎⁎ 0.018⁎⁎⁎ −0.087⁎⁎⁎ −0.113⁎⁎⁎ −0.118⁎⁎⁎
(0.91) (23.85) (5.64) (−11.61) (−190.57) (−47.51)
2007– 0.016 0.025⁎⁎⁎ 0.027⁎⁎⁎ −0.077⁎⁎⁎ −0.050⁎⁎⁎ −0.038⁎⁎⁎
(0.87) (18.11) (4.26) (−9.97) (−50.92) (−6.87)
Age (unknown) −0.008 −0.055⁎⁎⁎ −0.057⁎⁎⁎ −0.046⁎⁎⁎ −0.035⁎⁎⁎ −0.034⁎⁎⁎
(−0.74) (−95.94) (−24.43) (−7.93) (−76.24) (−18.27)
Terraced Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out
Detached 0.226⁎⁎⁎ 0.184⁎⁎⁎ 0.184⁎⁎⁎ 0.022⁎⁎⁎ −0.01⁎⁎⁎ −0.014⁎⁎⁎
(30.64) (471.26) (119.76) (5.89) (−37.86) (−11.69)
Semi-detached 0.104⁎⁎⁎ 0.088⁎⁎⁎ 0.088⁎⁎⁎ 0.006 0.004⁎⁎⁎ 0.003⁎⁎⁎
(17.25) (275.74) (72.37) (1.82) (14.60) (3.36)
Flat −0.029 −0.005⁎⁎⁎ −0.003 0.009 −0.035⁎⁎⁎ −0.039⁎⁎⁎
(−1.75) (−6.13) (−0.95) (1.21) (−55.30) (−14.86)
Freehold 0.050⁎⁎⁎ 0.061⁎⁎⁎ 0.064⁎⁎⁎ 0.021⁎⁎ 0.017⁎⁎⁎ 0.018⁎⁎⁎
(3.43) (88.84) (21.14) (3.12) (32.09) (8.03)
Deprivation index −0.010⁎⁎⁎ −0.009⁎⁎⁎ −0.009⁎⁎⁎ −0.0003 0.0005⁎⁎⁎ 0.001⁎⁎⁎
(−34.24) (−620.53) (−155.11) (−1.95) (45.98) (12.76)
Urban level Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out
Category 1
Urban level 0.107 −0.010 −0.028 −0.154⁎⁎⁎ 0.007 0.010
Category 2 (1.38) (−1.75) (−1.19) (−4.26) (1.73) (0.57)
Urban level 0.197⁎⁎ 0.081⁎⁎⁎ 0.074⁎⁎⁎ −0.149⁎⁎⁎ 0.055⁎⁎⁎ 0.060⁎⁎⁎
Category 3 (2.79) (15.60) (3.40) (−4.64) (13.23) (3.73)
Urban level 0.164 0.093⁎⁎⁎ 0.086⁎⁎⁎ −0.171⁎⁎⁎ 0.052⁎⁎⁎ 0.060⁎⁎
Category 4 (1.95) (15.87) (3.31) (−4.92) (11.33) (3.12)
Urban level 0.060 −0.050⁎⁎⁎ −0.069⁎⁎ −0.221⁎⁎⁎ −0.005 −0.01
Category 5 (0.76) (−9.28) (−3.10) (−6.01) (−1.23) (−0.59)
Urban level 0.059 −0.038⁎⁎⁎ −0.055⁎ −0.214⁎⁎⁎ −0.001 −0.004
Category 6 (0.74) (−7.01) (−2.48) (−5.81) (−0.16) (−0.22)
Urban level 0.107 0.034⁎⁎⁎ 0.018 −0.209⁎⁎⁎ 0.016⁎⁎⁎ 0.013
Category 7 (1.36) (6.29) (0.82) (−5.65) (3.90) (0.81)
Urban level 0.125 0.059⁎⁎⁎ 0.046⁎ −0.223⁎⁎⁎ 0.022⁎⁎⁎ 0.023
Category 8 (1.54) (10.80) (2.06) (−5.95) (5.40) (1.37)
Constant 7.753⁎⁎⁎ 7.959⁎⁎⁎ 7.958⁎⁎⁎ 0.245⁎⁎⁎ 0.029⁎⁎⁎ 0.034⁎
(93.27) (1455.60) (350.75) (6.35) (7.13) (2.04)
Quarterly fixed effects Yes Yes Yes
Postcode fixed effects Yes Yes Yes Yes Yes Yes
Adj. R2 0.681 0.994 0.689 0.251 0.994 0.786
N 18,135 62,601 306,915 18,130 95,830 306,826
t statistics based on robust standard errors are reported within parentheses.
⁎ p b 0.05.
⁎⁎ p b 0.01.
⁎⁎⁎ p b 0.001.
 F. Fuerst et al. / Energy Economics 48 (2015) 145–156 155

Table 7
Energy rating and price effect by region (dependent variable: log of price per square metre).

North East North West Yorkshire and Humberside East Midlands West Midlands East of England London South East South West

Average price (£/m2) 1558 1717 1755 1687 1891 2454 4079 2831 2301
EPC A/B 0.144⁎⁎⁎ 0.077⁎⁎⁎ 0.095⁎⁎⁎ 0.071⁎⁎⁎ 0.087⁎⁎⁎ 0.024⁎ −0.003 0.017 0.070⁎⁎⁎
(5.60) (4.74) (5.50) (3.66) (4.30) (2.21) (−0.16) (1.68) (5.26)
EPC C 0.027⁎⁎⁎ 0.028⁎⁎⁎ 0.016⁎⁎⁎ 0.028⁎⁎⁎ 0.029⁎⁎⁎ 0.015⁎⁎⁎ −0.006 0.015⁎⁎⁎ 0.021⁎⁎⁎
(4.51) (7.23) (3.86) (7.21) (7.07) (5.09) (−1.43) (6.22) (6.65)
EPC D Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out Hold-out
EPC E −0.025⁎⁎⁎ −0.016⁎⁎⁎ −0.020⁎⁎⁎ −0.021⁎⁎⁎ −0.011⁎⁎ 0.0002 −0.0004 0.003 −0.005
(−3.55) (−4.16) (−4.94) (−5.42) (−2.83) (0.08) (−0.12) (1.04) (−1.54)
EPC F −0.060⁎⁎⁎ −0.042⁎⁎⁎ −0.046⁎⁎⁎ −0.033⁎⁎⁎ −0.015 0.011 −0.010 0.029⁎⁎⁎ −0.012
(−3.41) (−5.02) (−5.29) (−4.32) (−1.89) (1.75) (−1.46) (5.40) (−1.81)

EPC G −0.220⁎⁎⁎ −0.162⁎⁎⁎ −0.119⁎⁎⁎ −0.071⁎⁎ −0.064⁎⁎⁎ −0.018 −0.107⁎⁎⁎ 0.005 −0.038⁎
(−5.35) (−7.70) (−6.38) (−3.03) (−3.41) (−1.03) (−5.51) (0.29) (−2.55)
Adj. R2 0.540 0.559 0.548 0.435 0.508 0.561 0.685 0.516 0.389
N 13,395 36,278 30,738 32,655 30,201 48,512 32,084 68,583 40,649

t statistics based on robust standard errors are reported within parentheses.

⁎ p b 0.05.
⁎⁎ p b 0.01.
⁎⁎⁎ p b 0.001.

rated D. Only eight out of 333,095 (twice) sold dwellings were in band
A. Given the careful approach taken to select the sample, there is no rea-
son to suspect that this breakdown of ratings is significantly different
from the population of transacted dwellings over the relevant time pe-
riod in this study. Not surprisingly, there is a clear relationship between
energy efficiency and age. Only 6% of dwellings built before 1900 had a
rating of C or better. The comparable figure for dwellings constructed
since 2007 is 94%. We find evidence of a positive association between
price per square metre and energy performance rating. We estimate
that, compared to dwellings rated D, those in bands A and B sold at a
premium of 5% and those in band C for a premium of 1.8%. Dwellings
rated F and E sold for approximately 1% less and dwellings rated G
sold for nearly 7% less. Apart from the dwellings in the highest EPC
band, the findings for price growth are broadly similar.
In order to put the percentage premiums into context, it is useful to
consider the average dwelling price in England and Wales. In August
2014 this was £177,824 according to Land Registry data.3 For the full
sample, dwellings in EPC bands A and B sold for a 5% premium, all else
equal; a monetary equivalent of approximately £8900 in relation to
the average house price. For dwellings in band C the premium was
1.8%, a monetary equivalent of approximately £3200. In comparison to
the cost of energy efficiency measures, these premiums seem broadly
plausible. For example, the combined cost of typical energy efficiency
measures that include loft and wall cavity insulation, boiler and space-
heating controls, and low energy lighting (but not renewable power
generation such as photovoltaic panels) would be approximately
£6100 for a two-bedroom Victorian end-of-terrace dwelling and
£7300 for a three-bedroom semi-detached dwelling built in the mid
twentieth century (BPF, 2013).
In the UK, as in most residential real estate markets, house price is
driven in the main by location, size and dwelling type. However, price
is also influenced by other attributes that determine the condition and
quality of the accommodation. Many of these attributes, such as
double-glazing, modern space and water heating systems, improve
energy efficiency and so it is not surprising that there is a positive rela-
tionship between price and the energy efficiency of a dwelling. We
should acknowledge that there are significant empirical challenges in
estimating the equilibrium prices of the component attributes in
house price models. For example, it is not practical to include the full
range of price determining variables in the model and, thus, hedonic
studies are afflicted by omitted variable and endogeneity problems. In
this study specifically, it is possible that dwellings with higher energy
3
https://www.gov.uk/government/uploads/system/uploads/attachment_data/file/
357080/HPIReport20140922.pdf.
ratings are also superior in terms of other attributes such as more mod-
ern kitchens and bathrooms, which may impact upon the results of our
cross-sectional estimation. However, by combining the cross-sectional
analysis with an augmented repeat-sales approach, a consistent pattern
emerges that is robust across model specifications, sub-groups and
time periods. Whilst the causality behind the energy efficiency price
premium may be subject to debate, this research has established the
existence of such a premium in the English housing market.
Acknowledgement
The authors wish to thank the UK Department of Energy and Climate
Change (TRN 297/11/2011) for financial support. Helpful comments on
earlier drafts were received from Chris Nicholls, Jeremy Moody and two
anonymous referees of this Journal. Franz Fuerst also wishes to ac-
knowledge the generous support of the Cambridge University Land So-
ciety (CULS) in enabling his research.
Appendix 1. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.eneco.2014.12.012.
References
Australian Bureau of Statistics, 2008. Energy efficiency rating and house prices in the ACT.
Report for Department of the Environment, Water, Heritage and Arts.
Bailey, M., Muth, R., Nurse, H., 1963. A regression method for real estate price index
construction. J. Am. Stat. Assoc. 58 (304), 933–942.
Bajari, P., Benkard, C., 2005. Demand estimation with heterogeneous consumers and unob-
served product characteristics: a hedonic approach. J. Polit. Econ. 113 (6), 1239–1276.
Bartik, T., 1987. The estimation of demand parameters in hedonic price models. J. Polit.
Econ. 95 (1), 81–88.
Blomquist, G., Berger, M., Hoehn, J., 1988. New estimates of quality of life in urban areas.
Am. Econ. Rev. 78, 89–107.
Bourassa, S., Hoesli, M., Sun, J., 2006. A simple alternative house price index method.
J. Hous. Econ. 15, 80–97.
BPF, 2013. Energy efficiency and the private rented sector, British Property Federation.
http://www.bpf.org.uk/en/files/event_related_documents/Energy_Efficiency_and_
the_Private_Rented_Sector.pdf.
Brounen, D., Kok, N., 2011. On the economics of energy labelling in the housing market.
J. Environ. Econ. Manag. 62, 166–179.
Cajias, M., Piazolo, D., 2013. Green performs better: energy efficiency and financial return
on buildings. J. Corp. Real Estate 15 (1), 53–72.
Case, B., Clapp, J., Dubin, R., Rodriguez, M., 2004. Modeling spatial and temporal house price
patterns: a comparison of four models. J. Real Estate Financ. Econ. 29 (2), 167–191.
Chernobai, A., Chernobai, E., 2013. Is selection bias inherent in housing transactions? An
equilibrium approach. Real Estate Econ. 41 (4), 887–924.
Costello, G., 2000. Pricing size effects in housing markets. J. Prop. Res. 17 (3), 203–219.
Costello, G., Watkins, C., 2002. Towards a system of local house price indices. Hous. Stud.
17, 857–873.
156 F. Fuerst et al. / Energy Economics 48 (2015) 145–156
Deng, Y., Li, Z., Quigley, J., 2011. Economic returns to energy-efficient investments in the
housing market: evidence from Singapore. Reg. Sci. Urban Econ. 42, 685–694.
Dinan, T., Miranowski, J., 1989. Estimating the implicit price of energy efficiency improve-
ments in the residential housing market: a hedonic approach. J. Urban Econ. 25 (1),
52–67.
Edlefsen, L., 1981. The comparative statics of hedonic price functions and other nonlinear
constraints. Econometrica 49, 1501–1520.
Eichholtz, P., Kok, N., Quigley, J., 2010. Doing well by doing good? Green office buildings.
Am. Econ. Rev. 100, 2492–2509.
Eichholtz, P., Kok, N., Quigley, J., 2013. The economics of green building. Rev. Econ. Stat. 95
(1), 50–63.
English Housing Survey, 2014. Headline report 2012–13. Department for Communities
and Local Government. https://www.gov.uk/government/uploads/system/uploads/
attachment_data/file/284648/English_Housing_Survey_Headline_Report_2012-13.
pdf.
Englund, P., Quigley, J., Redfearn, C., 1999. The choice of methodology for computing
housing price indexes: comparisons of temporal aggregation and sample definition.
J. Real Estate Financ. Econ. 19, 91–112.
Epple, D., 1987. Hedonic prices and implicit markets: estimating the demand and supply
functions for differentiated products. J. Polit. Econ. 95, 59–80.
Fuerst, F., McAllister, P., 2011a. Green noise or green value: measuring the price effects of
environmental certification in commercial buildings. Real Estate Econ. 39, 46–69.
Fuerst, F., McAllister, P., 2011b. Eco-labeling in commercial real office markets: do LEED
and Energy Star offices obtain multiple premiums? Ecol. Econ. 70, 1220–1230.
Fuerst, F., McAllister, P., 2011c. The impact of energy performance certificates on the rent-
al and capital values of commercial property assets. Energy Policy 39, 6608–6614.
Gatzlaff, D., Haurin, D., 1997. Sample selection bias and repeat-sales index estimates.
J. Real Estate Financ. Econ. 14, 33–50.
Giles, D., 2011. Interpreting Dummy Variables in Semi-logarithmic Regression Models:
Exact Distributional Results (Working Paper). University of Victoria, Department of
Economics, Victoria, B.C.
Halvorsen, R., Palmquist, R., 1980. The interpretation of dummy variables in semilogarith-
mic equations. Am. Econ. Rev. 70 (3), 474.
Halvorsen, R., Pollakowski, H., 1981. The effects of fuel prices on house prices. Urban Stud.
18, 205–211.
Hyland, M., Lyons, R., Lyons, S., 2013. The value of domestic building energy efficiency:
evidence from Ireland. Energy Econ. 40, 943–952.
Jansen, S., de Vries, P., Coolen, H., Lamain, C., Boelhouwer, P., 2008. Developing a house
price index for The Netherlands: a practical application of weighted repeat sales.
J. Real Estate Financ. Econ. 37, 163–186.
Johnson, R., Kaserman, D., 1983. Housing market capitalization of energy‐saving durable
good investments. Econ. Inq. 21 (3), 374–386.
Kok, N., Jennen, M., 2012. The impact of energy labels and accessibility on office rents.
Energy Policy 46, 489–497.
Laquatra, J., 1986. Housing market capitalization of thermal integrity. Energy Econ. 8 (3),
134–138.
Laquatra, J., Dacquisto, D., Emrath, P., Laitner, J., 2002. Housing market capitalization of
energy efficiency revisited. ACEEE Summer Study of Energy Efficiency in Buildings,
Teaming for Efficiency, Proceedings. 8.
Malpezzi, S., 2003. Hedonic Pricing Models: A Selective and Applied Review, in Housing
Economics and Public Policy: Essays in Honor of Duncan MacLennan. In: Sullivan,
T.O., Gibbs, K. (Eds.), Blackwell.
Quigley, J., 1984. The production of housing services and the derived demand for residen-
tial energy. RAND J. Econ. 555–567.
Quigley, J., Rubinfeld, D., 1989. Unobservables in consumer choice: residential energy and
the demand for comfort. Rev. Econ. Stat. 416–425.
Reichardt, A., Fuerst, F., Rottke, N., Zietz, J., 2011. The business case of sustainable building
certification: a panel data approach. J. Real Estate Res. 34, 99–126.
Roback, J., 1982. Wages, rents, and the quality of life. J. Polit. Econ. 90, 1257–1278.
Rosen, S., 1974. Hedonic prices and explicit markets: production differentiation in pure
competition. J. Polit. Econ. 82, 34–55.
Van Garderen, K.J., Shah, C., 2002. Exact interpretation of dummy variables in semiloga-
rithmic equations. Econ. J. 5, 149–159.
Wiley, J., Benefield, J., Johnson, K., 2010. Green design and the market for commercial
office space. J. Real Estate Financ. Econ. 41, 228–243.