# RECORD 2bbd25e60ca8ea2625da98d69a5096e00200229f
# TITLE: Does energy efficiency matter for prices of tenant-owned apartments?
# YEAR: 2022.0  VENUE: Environmental science and pollution research international  DOI: 10.1007/s11356-022-20482-w
# AUTHORS: David Stenvall; P. Cerin; Bo Sjö; Gazi Salah Uddin
# PAGES: 15  CANDIDATE TABLE PAGES (1-based): [8, 9, 10, 11, 12]
# ABSTRACT: In this study, we use Energy Performance Certificate (EPC) information to investigate the effect of energy efficiency on the selling price of Swedish tenant-owned apartments. While there is a large body of literature on how energy efficiency affects the sales price of single-family houses, none has exclusively focused on tenant-owned apartments. For owners of tenant-owned apartments in Sweden, heating is for a large share included in the monthly fee paid to the tenant association, which usually does not change on a short-term basis. This raises the question whether homebuyers’ incentives for acquiring energy-efficient tenant-owned apartments are large enough to be capitalized into the prices. By hedonic models and matching methods, we found mixed results. In our most optimistic scenarios, tenant-owned apartments enclosed in energy-efficient buildings are sold with a premium of approximately 0.8 to 1.2% compared to apartments in non-efficient buildings. The results in this study are not robust to all model specifications and vary across regions. In comparison with recent studies using data for single-family houses in Sweden, our detected capitalization is smaller. Our results highlight a need for targeted measures if EPCs are to be fully capitalized in prices for all dwelling types in which buyers have different economic incentives for reducing their energy consumption.


===== PAGE 1 (score 9, numeric-paren 0, table-words 0, var-words 9) =====

Vol.:(0123456789)
1 3
https://doi.org/10.1007/s11356-022-20482-w
RESEARCH ARTICLE
Does energy efficiency matter for prices of tenant‑owned apartments?
David Stenvall1 · Pontus Cerin1 · Bo Sjö1 · Gazi Salah Uddin1
Received: 15 July 2021 / Accepted: 23 April 2022 
© The Author(s) 2022
Abstract
In this study, we use Energy Performance Certificate (EPC) information to investigate the effect of energy efficiency on 
the selling price of Swedish tenant-owned apartments. While there is a large body of literature on how energy efficiency 
affects the sales price of single-family houses, none has exclusively focused on tenant-owned apartments. For owners of 
tenant-owned apartments in Sweden, heating is for a large share included in the monthly fee paid to the tenant association, 
which usually does not change on a short-term basis. This raises the question whether homebuyers’ incentives for acquiring 
energy-efficient tenant-owned apartments are large enough to be capitalized into the prices. By hedonic models and matching 
methods, we found mixed results. In our most optimistic scenarios, tenant-owned apartments enclosed in energy-efficient 
buildings are sold with a premium of approximately 0.8 to 1.2% compared to apartments in non-efficient buildings. The 
results in this study are not robust to all model specifications and vary across regions. In comparison with recent studies 
using data for single-family houses in Sweden, our detected capitalization is smaller. Our results highlight a need for targeted 
measures if EPCs are to be fully capitalized in prices for all dwelling types in which buyers have different economic incen-
tives for reducing their energy consumption.
Keywords  Energy performance certificates · Housing markets · Tenant-owned apartments · Energy efficiency · Green 
premium
JEL Classification  D10 · Q51 · R20
Introduction
Reducing the speed of climate change has in the recent dec-
ade become as a top priority for numerous governments, 
scientists, and institutions. The EU has set up ambitious 
GHG emission targets in a vision for 2050, which could 
pose one important step-stone for the world to reach below 
the Paris 2-degree target and, as the union puts it, keep it 
down to 1.5 °C. This will, consequently, put a lot of pressure 
on the housing sector to abate its considerable large share of 
­CO2 emissions compared to other sectors. In this context, a 
common European approach to reduce the energy usage by 
dwellings was introduced through the Energy Performance 
of Buildings Directive (EPBD 2002/91/EC). The directive 
includes a certification scheme aimed towards dwellings, 
namely, the Energy Performance Certificates (EPCs).1 
One of the main purposes of the introduction of EPCs was 
to decrease potential information asymmetries between 
owners and prospective buyers with regard to the energy 
performance of dwellings. Hence, these certifications are 
supposed to make potential buyers aware of the energy per-
formance of their future homes or buildings (e.g., Brounen 
and Kok 2011; Bio Intelligence Service 2013; Cerin et al. 
2014; Fuerst et al. 2015; Fuerst et al. 2016a, b; EC 2020). 
Responsible Editor: Philippe Garrigues
 *	 Gazi Salah Uddin 
	
gazi.salah.uddin@liu.se
	
David Stenvall 
	
david.stenvall@liu.se
	
Pontus Cerin 
	
pontus.cerin@liu.se
	
Bo Sjö 
	
bo.sjo@liu.se
1	
Department of Management and Engineering, Linköping 
University, Linkoping, Sweden
1  For more information regarding the EPC in EU, see e.g., EC 
(2020). In the Swedish case, more information can be found on the 
website of the National Board of Housing, Building and Planning 
(Boverket 2019).
/ Published online: 4 May 2022
Environmental Science and Pollution Research (2022) 29:66793–66807


===== PAGE 8 (score 103, numeric-paren 73, table-words 5, var-words 11) =====

1 3
the number of rooms according to the VIF values (variance 
inflation factors). As both are control variables, they are kept 
in the estimations. The VIF value for our treatment vari-
able for energy-efficient tenant-owned apartments, HIGH, 
is very low and does not suggest any problems with multi-
collinearity. Our R2 over 81% is notably high, but common 
for hedonic price studies on the housing market, especially 
given that our sample consists of apartments which are more 
homogeneous compared to houses. As HIGH is insignifi-
cant, we do not find any effect of energy efficiency on the 
sales price of the tenant-owned apartments.
In column (2), we instead use the 4-digit postcode to 
control for more detailed locational effects. Our R2 now 
increases to 93%, which indicates that a more detailed con-
trol for location is important in explaining the sales price. 
Our treatment, HIGH, is now significant and indicates a 
price premium of approximately 1.1% for energy-efficient 
tenant-owned apartments. This can be interpreted as approx-
imately 11,000 SEK more for every 1 million in sales price 
for the energy-efficient homes, ceteris paribus, compared to 
the reference category (homes enclosed in buildings with 
higher energy performance than 75 kWh/m2). The size of 
the premium is smaller compared with the findings of Cerin 
et al. (2014) and Wilhelmsson (2019) that found larger pre-
miums when using data for houses rather than apartments. 
As emphasized in “Introduction,” a large share of the home-
owners of tenant-owned apartments has heating included in 
the monthly fee paid to the association each month, which 
usually does not change on a short-term basis. The fee is not 
based on the usages per apartment but rather the joint costs 
for the association. That the premium is quite low might 
simply reflect low economic incentives for many buyers of 
tenant-owned apartments. In column (3), we include our 
dummy, LOW, for tenant-owned apartments enclosed in 
tenant-owned buildings with low energy efficiency (hereafter 
denoted as low-efficient apartments). By including dummies 
for apartments with both high and low energy efficiency, the 
reference category is now apartments in between these two 
categories (medium energy efficient). The insignificance of 
LOW indicates that low-efficiency apartments are not sold 
Table 2   Full sample — hedonic price estimations
The dependent variable is the logarithmic sales price. Standard errors in parenthesis. Columns (1), (2), and (3) are estimated with robust stand-
ard errors. Postcode fixed effects is based on 4 digits if nothing else is stated. HIGH is a dummy variable for tenant-owned apartments enclosed 
in tenant-owned buildings with high energy efficiency; LOW is a dummy for tenant-owned apartments enclosed in tenant-owned buildings with 
low energy efficiency; Area is the size of the dwelling in square meters (log); Room is the number of rooms; Building year (age) is a dummy 
equal to 1 if the apartment was constructed during the given range of years (omitted category is pre-1901). Rent is the fee paid to the tenant asso-
ciation each month in SEK (log). We also include a dummy variable equal to 1 for apartments with double construction years (see the explana-
tion in the data section), for which the result is not shown
Statistical significance is denoted by ***p < 0.01, **p < 0.05, *p < 0.1
(1)
(2)
(3)
(4)
(5)
Energy efficiency (EPC)
  HIGH
0.001 (0.007)
0.011** (0.005)
0.011** (0.005)
0.007* (0.004)
0.007* (0.004)
  LOW
 − 0.001 (0.002)
 − 0.001 (0.002)
Dwelling-specific controls
  Area
0.862*** (0.019)
0.746*** (0.012)
0.746*** (0.012)
0.751*** (0.006)
0.751*** (0.006)
  Rent
 − 0.277*** (0.023) 
 − 0.144*** (0.014)
 − 0.143*** (0.014)
 − 0.156*** (0.004)
 − 0.156*** (0.004)
  Room
0.049*** (0.004)
0.060*** (0.002)
0.060*** (0.002)
0.058*** (0.002)
0.058*** (0.002)
  Building year
    1901–1921
 − 0.041***
 − 0.002
 − 0.002
 − 0.006
 − 0.006
    1922–1941
 − 0.143***
 − 0.066***
 − 0.066***
 − 0.070***
 − 0.070***
    1942–1961
 − 0.280***
 − 0.116***
 − 0.116***
 − 0.120***
 − 0.120***
    1962–1976
 − 0.467***
 − 0.165***
 − 0.165***
 − 0.163***
 − 0.163***
    1977–1991
 − 0.304***
 − 0.143***
 − 0.144***
 − 0.139***
 − 0.139***
    1992–2005
 − 0.218***
 − 0.074***
 − 0.074***
 − 0.063***
 − 0.063***
    2006–2019
 − 0.123***
 − 0.001
 − 0.002
 − 0.004
 − 0.004
Constant
13.48***
13.48***
13.48***
13.54***
13.54***
Time fixed effects
Yes
Yes
Yes
Yes
Yes
Robust regression
No
No
No
Yes
Yes
Postal code fixed effects
(2 digits)
Yes
Yes
Yes
Yes
R2
0.812
0.933
0.933
0.943
0.943
Obs
21,696
21,696
21,696
21,695
21,696
66800
Environmental Science and Pollution Research (2022) 29:66793–66807


===== PAGE 9 (score 51, numeric-paren 3, table-words 18, var-words 11) =====

1 3
with a discount compared to medium-efficiency homes all 
else equal. In columns (4) and (5), we try to mitigate issues 
related to potential outliers by performing the estimations 
using the robust regression. The effect of high energy effi-
ciency on the sales price is now reduced to a premium of 
approximately 0.7% that now is weakly significant. The 
effect of low efficiency on prices is still insignificant.
In Table 3, we present the result from regional subsam-
ples. The category South consists of the observations from 
the municipality of Malmö while the Mid category is based 
on the sales from Gothenburg and Linköping. As only a few 
observations of the sample from the northern cities Luleå 
and Umeå do qualify as energy efficient according to our 
boundaries, we do not run estimation solely based on these 
observations. As observed, the coefficient for HIGH is larg-
est in the southern sample followed by the Stockholm. The 
effect is, however, not statistically significant for any region.
Table 4 shows the hedonic price estimations using the 
sample of tenant-owned apartments enclosed in a tenant-
owned building with an EPC grade. In (1), we classify 
energy-efficient apartments as those enclosed in a build-
ing with the grades A or B and the other labels (C–G) as 
the reference category. Our results indicate that the effect 
of energy efficiency is positively related to the sales price 
but only weakly significant. The estimated premium is 
approximately 1.5%, all else equal, which is higher com-
pared with the effect of energy efficiency in the full sample 
in Table 2. This 1.5% premium can be compared with the 
ten-basis point reduction in mortgage rate that buyers can 
get by certain banks if they acquire a home with an energy 
grade of A or B. Assuming an annual mortgage rate of 2% 
on a 4 million SEK loan, a 0.10-percentage point reduction 
is approximately 4000 SEK less per year. If not considering 
the time value of money or any other potential cost savings 
from owning an energy-efficient home, a 1.5% premium is 
paid off in 15 years based solely on the mortgage rate reduc-
tion. Therefore, a 1.5% premium is not unreasonable based 
on these economic incentives. In Table 4, column (2), we 
instead define the energy-efficient labels as A–C and com-
pare it with the less efficient grades (D–G). Even though 
the coefficient for the energy efficient grades is positive, the 
effect is not significant. Hence, our results from (1) and (2) 
suggest that the price premium is only present in the most 
energy-efficient labels, A and B. That these two labels also 
provide the largest interest rate reduction for homebuyers 
with green loans might perhaps be a part of the explana-
tion. In column (3), we add dummies for the least energy 
efficient labels (E–G). Our result from this estimation shows 
no significant discount compared with the reference category 
(label D). In (4) and (5), we run the estimations using robust 
regression. Our effect found in column (4) indicates a 1% 
premium for labels A and B. However, this is not enough 
to reach statistical significance. While the robust regression 
barely drops any observations, it still weights the observa-
tions that are not dropped. Hence, the reduced effect from 
1.5% premium in column (1) to 1% in column (4) is likely 
due to the weighting of potential outliers rather than the 
actual dropping of observations.
Robustness check: neighborhood controls 
and matching
As one of the few Swedish studies, we also control for some 
neighborhood characteristics in our hedonic price models. 
In this section, we use the full sample from Table 2. Because 
the neighborhood data is based on DeSO regions, observa-
tions belonging to the same region will get the same neigh-
borhood values. In Table 5, columns (1) and (2) we start the 
investigation by dividing the data into income categories 
based on the income data. We calculate the income quar-
tiles in the sample based on the incomes attached to each 
observation. Apartments with incomes below quartile 1 is 
denoted by IC1, apartments with incomes equal to or larger 
than quartile 1 but smaller than quartile 2 is denoted by IC2, 
and, lastly, apartments with incomes equal to or larger than 
quartile 2 but smaller than quartile 3 is denoted by IC3.
As observed in Table 5, column (1), the income dum-
mies (IC 1–3) are all significant with negative coefficients. 
This means that there is a significant difference in price, all 
else equal, between observations within our included income 
categories and our omitted reference group (high incomes 
above the third quartile). Our energy efficiency variable, 
Table 3   Regression with regional subsamples from the full sample
The dependent variable is the logarithmic sales price. Standard errors 
are in parenthesis. Postcode fixed effects are based on 4 digits. Robust 
standard errors are used in the estimations. The mid region contains 
dwellings from the municipalities Linköping and Gothenburg, and the 
southern region consists of apartments from the Malmö municipal-
ity. HIGH is a dummy for tenant-owned apartments enclosed in ten-
ant-owned buildings with high energy efficiency. Dwelling-specific 
variables include the following: area (log), number of rooms in each 
apartment, rent or fee paid to the tenant association each month (log), 
and dummy for construction year (same as Tables 2 and 3). We also 
include a dummy variable equal to 1 for apartments with double con-
struction years (see the explanation in the data section)
Statistical significance is denoted by ***p < 0.01, ** p < 0.05, * 
p < 0.1
Stockholm
South
Mid
HIGH
0.004 (0.006) 0.015 (0.017) 0.002 (0.008)
Dwelling-specific 
controls
Yes
Yes
Yes
Time fixed effects
Yes
Yes
Yes
Postal code fixed 
effects
Yes
Yes
Yes
R2
0.913
0.885
0.912
Obs
11 468
4142
5285
66801
Environmental Science and Pollution Research (2022) 29:66793–66807


===== PAGE 10 (score 100, numeric-paren 68, table-words 7, var-words 9) =====

1 3
HIGH, indicates a price premium of approximately 0.8%, 
but it is only of a weakly statistically significant size. How-
ever, when using 3-digit postcodes (a wider geographical 
area) in column (3), controlling for outliers using robust 
regression in (2) and (5), including a full set of neighbor-
hood controls (3)–(5), our energy efficiency variable HIGH 
is no longer significant.
In Table 6, we test if our estimations vary depending 
on the boundaries for which we classify the apartments as 
energy efficient. As observed, the coefficient when narrow-
ing the boundaries, columns in (1) estimate of HIGH, is 
larger compared to the results in Table 5, but still not sig-
nificant. This might be a sample size problem with relatively 
few observations reaching energy efficiency status. If instead 
increasing the boundaries in (2), we observe a non-signifi-
cant premium for energy-efficient tenant-owned apartments 
of approximately 0.5%. In (3), we include an interaction 
between our energy efficiency measure and the net income 
in the DeSO in which each observation belongs to. By this, 
we want to analyze if the effect of energy efficiency on prices 
is different if the energy-efficient apartments are sold in a 
low- versus high-income area. As a comparison, an energy-
efficient apartment in a neighborhood with a net income of 
250,000 SEK (which is slightly below median in the sample) 
will according to our estimates be sold with a premium of 
1.77% compared to a non-efficient apartment. However, in 
a neighborhood with a net income of 320,000 SEK (above 
median), an energy-efficient apartment is sold with a dis-
count of − 0.2%. The interaction between net income and 
HIGH is significant even if we run the estimation with our 
standard boundary of 75 kWh/sqm. One possible explana-
tion is that there might be a larger share of apartments in 
which heating is not included in the monthly fee in low- 
versus high-income areas. In such case, the economic incen-
tives might be lower for high-income individuals. It could 
also be that high-income individuals simply do afford to not 
Table 4   Grade sample — hedonic price estimations
The dependent variable is the logarithmic tenant-owned apartment sales price. Standard errors in parenthesis. Columns (1), (2), and (3) are 
estimated with robust standard errors. Postcode fixed effects is based on 4 digits if nothing else is stated. Area is the of the size of the dwelling 
in sqm (log); Room is the number of rooms; Building year (age) is a dummy equal 1 if the apartment was constructed during the given range of 
years (omitted category is pre 1901). Rent is the fee paid to the tenant association each month in SEK (log); Energy grades are dummy variables 
equal to 1 if the tenant-owned apartment belongs to a tenant-owned building with the specified EPC-grade/grades. In columns (3) and (5), the 
energy grade D is the omitted category, while grades C–G is omitted in columns (1) and (4) and D–G in (2). We also include a dummy variable 
equal to 1 for apartments with double construction years (see explanation in the data section), for which the result is not shown. Columns (1), 
(2), and (3) are ordinary hedonic price estimations while (4) and (5) are hedonic price model estimated by robust regressions
We denote statistical significance by ***p < 0.01, **p < 0.05, *p < 0.1
(1)
(2)
(3)
(4)
(5)
Energy grade (EPC)
  AB
0.015* (0.009)
0.010 (0.008)
  ABC
0.001 (0.005)
0.002 (0.006)
 − 0.001 (0.006)
  EFG
0.002 (0.005)
 − 0.002 (0.005)
Dwelling-specific controls
  Area
0.749*** (0.023)
0.750*** (0.023)
0.750*** (0.023)
0.750*** (0.012)
0.750*** (0.012)
  Rent
 − 0.164*** (0.025)
 − 0.165*** (0.025)
 − 0.165*** (0.025)
 − 0.170*** (0.009)
 − 0.170*** (0.009)
  Room
0.065*** (0.004)
0.065*** (0.004)
0.065*** (0.004)
0.059*** (0.003)
0.059*** (0.003)
  Building year
    1901–1921
0.000
0.000
0.000
0.000
 − 0.001
    1922–1941
 − 0.059***
 − 0.059***
 − 0.060***
 − 0.068***
 − 0.068***
    1942–1961
 − 0.129***
 − 0.129***
 − 0.129***
 − 0.136***
 − 0.136***
    1962–1976
 − 0.172***
 − 0.172***
 − 0.172***
 − 0.165***
 − 0.165***
    1977–1991
 − 0.166***
 − 0.166***
 − 0.166***
 − 0.158***
 − 0.158***
    1992–2005
 − 0.091***
 − 0.091***
 − 0.091***
 − 0.079***
 − 0.079***
    2006–2019
0.016
0.019
0.019
0.013
0.014
Constant
13.80***
13.80***
13.80***
13.97***
13.97***
Time fixed effects
Yes
Yes
Yes
Yes
Yes
Robust regression
No
No
No
Yes
Yes
Postal code fixed effects
Yes
Yes
Yes
Yes
Yes
R2
0.936
0.936
0.936
0.945
0.945
Obs
6555
6555
6555
6555
6553
66802
Environmental Science and Pollution Research (2022) 29:66793–66807


===== PAGE 11 (score 49, numeric-paren 18, table-words 9, var-words 10) =====

1 3
care about energy savings. As we do not have information 
on heating included for the full sample, we are not able to 
control for this.
As described in the data section, ordinary hedonic price 
estimations might, however, suffer from methodical prob-
lems. Using a propensity score method, we can address the 
issue of differences between the group with high energy 
efficiency (treatment) versus the non-energy efficient group 
(control). In Table 8, we use two different methods of pro-
pensity score. First, we estimate the propensity score using 
a probit model and include the score in the hedonic model 
as a covariate. The propensity score in the first step is esti-
mated using the treatment variables (HIGH and AB) as the 
dependent variable and the dwelling- and location-specific 
Table 5   Full sample — hedonic 
prices with neighborhood 
controls
The dependent variable is the logarithmic tenant-owned apartment sales price. Standard errors in paren-
thesis. HIGH is a dummy variable for tenant-owned apartments enclosed in tenant-owned buildings with 
high energy efficiency. IC1 is a dummy for observations that belong to a postcode with low incomes. IC2 
is a dummy for low to median income postcodes while IC3 is a dummy for median to upper incomes. 
Neighborhood controls consists of median net income (log), share of persons with foreign origin, share 
of persons in rental apartments, and share of persons receiving financial aid and number of households 
(log). Dwelling-specific variables are area (log), number of rooms in each apartment, rent or fee paid to the 
tenant-association each month (log), and dummy for construction year (same as Tables 2 and 4). We also 
include a dummy variable equal to 1 for apartments with double construction years (see explanation in the 
data section)
We denote statistical significance by ***p < 0.01, **p < 0.05, *p < 0.1
(1)
(2)
(3)
(4)
(5)
HIGH
0.008* (0.005)
0.006 (0.004)
 − 0.003(0.005)
0.003 (0.005)
0.000 (0.004)
IC1
 − 0.113***
 − 0.092***
IC2
 − 0.053***
 − 0.038***
IC3
 − 0.022***
 − 0.015***
Dwelling-specific controls
Yes
Yes
Yes
Yes
Yes
Time fixed effects
Yes
Yes
Yes
Yes
Yes
Robust regression
No
Yes
No
No
Yes
Postal code fixed effects
(4 digits)
(4 digits)
(3 digits)
(4 digits)
(4 digits)
Neighborhood controls
No
No
Yes
Yes
Yes
R2
0.935
0.944
0.924
0.938
0.946
Obs
21,696
21,691
21,696
21,696
21,696
Table 6   Robustness and 
alternative boundaries for 
energy efficiency
The dependent variable is the logarithmic tenant-owned apartment sales price. Standard errors are in 
parenthesis. HIGH is a dummy for tenant-owned apartments enclosed in tenant-owned buildings with high 
energy efficiency. Dwelling-specific variables consist of the following: area (log), number of rooms in each 
apartment, rent or fee paid to the tenant-association each month (log), and dummy for construction year 
(same as Tables 2 and 4). Neighborhood controls include median net income (log), share of persons with 
foreign origin, share of persons in rental apartments, share of persons receiving financial aid, and number 
of households (log). We also include a dummy variable equal to 1 for apartments with double construction 
years (see explanation in the data section)
Statistical significance is denoted by ***p < 0.01, **p < 0.05, *p < 0.1
(1)
(2)
(3)
HIGH (≤ 50 kWh/m2)
0.010 (0.012)
HIGH (≤ 100 kWh/m2)
0.005 (0.003)
0.468*** (0.098)
HIGH (≤ 100 kWh/m2) × net income
 − 0.082*** (0.017)
Neighborhood controls
Yes
Yes
Yes
Dwelling-specific controls
Yes
Yes
Yes
Robust regression
No
No
No
Time fixed effects
Yes
Yes
Yes
Postal code fixed effects
(4 digits)
(4 digits)
(4 digits)
R2
0.938
0.938
0.938
Obs
21,696
21,696
21,696
66803
Environmental Science and Pollution Research (2022) 29:66793–66807


===== PAGE 12 (score 36, numeric-paren 6, table-words 11, var-words 7) =====

1 3
variables from the hedonic estimations as independent vari-
ables. Here, we chose to only include dummies with high 
energy efficiency and not for low efficiency. In the second 
method, we match the treatment and control groups based 
on the estimated propensity score using a nearest neighbor 
approach. We thereafter run the hedonic regression on this 
matched sample using weights from the matching process. 
For a discussion of these methods, see, e.g., Stuart (2010).
Table 7 shows the means in the sales price and covari-
ates before and after propensity score matching. As sus-
pected, we can observe that the treatment group in the 
unmatched samples largely differs regarding the age of 
the tenant-owned apartments. The control groups include 
older apartments that are both smaller and cheaper on 
average. However, after the matching procedure these dif-
ferences have been largely reduced. The balance in the 
matched samples is tested by t-tests and mean standardized 
differences, which can be found in appendix Table A3. 
These tests show that there is no significant difference 
between the control and treatment groups after match-
ing. Table 8 shows the results from the propensity score 
estimations.
Table 7   Mean in covariates 
before and after propensity 
score matching
This table shows the treatment and control groups before and after propensity score matching. The treat-
ment in the full sample is the variable HIGH, which is a dummy for tenant-owned apartments enclosed in 
tenant-owned buildings with high energy efficiency. The treatment for the graded sample is dummy vari-
ables equal to 1 if the tenant-owned apartment belongs to a tenant-owned building with the energy grade A 
or B. The matched full sample is equivalent to the sample used in estimations of column (2) in Table 8. For 
the graded matched sample, this is equivalent to the data used in estimations of column (4) in Table 8
Full sample
Graded sample
Unmatched
Matched
Unmatched
Matched
Treatment
Control
Treatment
Control
Treatment
Control
Treatment
Control
Price
3,742,086
3,338,207
3,761,089
3,754,787
3,605,326
3,129,489
3,697,648
3,624,645
Area
70.9
63.6
71.1
72.5
70.2
65.9
70.1
69.0
Age
26.5
66.6
26.8
28.8
19.27
53.6
20.0
23.0
Rent
3800.4
3495.4
3805.1
3975.4
3841.6
3637.1
3839.4
3814.1
Room
2.7
2.4
2.7
2.8
2.7
2.5
2.7
2.6
Obs
1397
20,299
1377
1377
424
6131
394
394
Table 8   Result after propensity 
score estimations
This table shows the hedonic price model with estimated propensity score as a covariate columns (1) and 
(3) as well as on propensity score matched samples (columns (2) and (4)). The dependent variable is the 
logarithmic tenant-owned apartment sales price. Standard errors are in parenthesis. HIGH is a dummy for 
tenant-owned apartments enclosed in tenant-owned buildings with high energy efficiency. Postcode fixed 
effects are 4 digits if nothing else is stated. The treatment for the graded sample is dummy variables equal 
to 1 if the tenant-owned apartment belongs to a tenant-owned building with the energy grade A or B. 
Dwelling-specific variables include area (log), number of rooms in each apartment, rent or fee paid to the 
tenant-association each month (log), and dummy for construction year (same as Tables 2 and 3). We also 
include a dummy variable equal to 1 for apartments with double construction years (see explanation in the 
data section)
Statistical significance is denoted by ***p < 0.01, **p < 0.05, *p < 0.1
Full sample
Graded sample
(1)
(2)
(3)
(4)
HIGH
0.009* (0.005)
0.012** (0.005)
AB
0.011 (0.009)
0.014 (0.009)
Dwelling-specific controls
Yes
Yes
Yes
Yes
Time fixed effects
Yes
Yes
Yes
Yes
Postal code fixed effects
Yes
Yes
Yes
Yes
Regression on matched sample
No
Yes
No
Yes
Propensity score as control
Yes
No
Yes
No
R2
0.927
0.911
0.929
0.939
Obs
16,511
2754
3638
788
66804
Environmental Science and Pollution Research (2022) 29:66793–66807


===== PAGE 13 (score 42, numeric-paren 2, table-words 16, var-words 7) =====

1 3
As we can see from columns (1) and (3), including the 
propensity score as a covariate largely reduces the effect of 
high energy efficiency on the sales price of tenant-owned 
apartments. The coefficient is positive and ranges from a 0.9 
to 1.1% premium, ceteris paribus, for tenant-owned apart-
ments enclosed in energy-efficient tenant-owned buildings 
compared to non-efficient. While the effect in the full sample 
is weakly significant, the estimated premium in the graded 
sample is insignificant. Looking at the matched samples 
(columns (2) and (4)), the effect is similar with the coef-
ficient ranging from 1.2 to 1.4% price premium. As before, 
the estimations based on the full sample is weakly significant 
but the effect in the graded sample is insignificant. Regard-
ing the insignificance in column (4), it is possible that the 
lack of significance could be a power issue considering the 
few observations. The coefficient for the treatment variable 
(AB) does not change much compared to the hedonic price 
model estimates in Table 4, column (1). We also observe 
that the coefficient for energy efficiency in Table 8, column 
(2), i.e., the non-graded sample, decreases compared to its 
corresponding hedonic estimates in Table 2. In this sample, 
we do not have the same issue with power, which indicates 
that other problems (such as non-randomness or sample 
selection) potentially may explain the differences between 
Tables 2 and 8.
Lastly, we use an alternative matching method, the 
coarsened exact matching (CEM) by Iacus et al. (2012). 
The results of the CEM are dependent on the variables cho-
sen for the matching procedure and the cut points through 
which these variables are coarsened. We chose to match on 
the most important control variables age, log rent, log area, 
and the number of rooms in each tenant-owned apartment. 
While running the CEM in Stata, it provides an algorithm 
to automatically choose the cut points. However, we chose 
the cut points by our own to ensure that this is done based 
on theoretical knowledge.9 We test different specifications 
before we end up with the final cut points. To quantify how 
much our cut points reduce the imbalances between the 
control and treatment observations, we use the multivariate 
L1 distance. The goal is to produce a model with as low a 
distance as possible. The distance in our chosen models can 
be found in Table A4, which are lower compared with the 
multivariate L1 distance produced by the Stata algorithm. 
In accordance with the procedure of Blackwell et al. (2009), 
we add the matched control variables to the hedonic regres-
sion to control for the remaining imbalances. We also add 
the control variables that were not used in the matching. Our 
results are shown in Table 9.
As observed, the results from the CEM are lower com-
pared to the estimates produced from the propensity score or 
hedonic estimation in Tables 2 and 4. The estimated effect 
for the graded or full sample is not statistically significant. 
The overall result from the matching procedures gives 
some conflicting evidence. Although the propensity score 
approach in Table 7 indicates price premiums of energy effi-
ciency close to the hedonic price model in Tables 2 and 4, 
the premiums indicated by the CEM are much lower. While 
results from the CEM are sensitive to the chosen cut points, 
the propensity score estimates are sensitive to which control 
observations the treatments are matched to. It is therefore 
very difficult to conclude which of these results are the most 
credible. In the previous literature, Wilhelmsson (2019) 
found that the price premium for single-family houses was 
substantially reduced when controlling for selection bias 
using propensity score matching. Against this background, 
we are not able to rule out that the results from the hedonic 
price models in Tables 2 and 4, at least to some extent, could 
be affected by sample selection bias. If such bias exists, our 
results from this section show that the price premium for 
energy-efficient tenant-owned apartments is lower compared 
with the results presented in Tables 2 and 4.
Table 9   Hedonic estimations after coarsened exact matching (CEM)
This table shows the results of hedonic price estimations weighted on 
the matched dataset after CEM. Standard errors are in parenthesis. 
The dependent variable is the logarithmic tenant-owned apartment 
sales price. HIGH is a dummy for tenant-owned apartments enclosed 
in tenant-owned buildings with high energy efficiency. Postcode fixed 
effects are based on 4 digits if nothing else is stated. The treatment 
for the graded sample is dummy variables equal to 1 if the tenant-
owned apartment belongs to a tenant-owned building with the energy 
grade A or B. Dwelling-specific variables include area (log), number 
of rooms in each apartment, rent or fee paid to the tenant-association 
each month (log), and dummy for construction year (same as Tables 2 
and 3). We also include a dummy variable equal to 1 for apartments 
with double construction years (see explanation in the data section)
Full sample
Graded sample
HIGH
0.005 (0.005)
AB
0.007 (0.009)
Dwelling-specific controls
Yes
Yes
Time fixed effects
Yes
Yes
Postal code fixed effects
Yes
Yes
R2
0.931
0.933
Obs
17,702
4373
9  Our cut points are as follows: age (0 14.5 29.5 44.5 59.5 79.5 99.5 
120.5), lrent (0 7.7 8 8.3 8.45 8.52 8.7 10), larea (0 3.1 3.45 3.8 4 
4.2 4.5 4.75 5 6), and rooms (0 2.25 3.75 5.25 10). For the cut points 
of the variables with a logarithmic transformation (larea and lrent), 
these are chosen with respect to corresponding values for the non-
logged variables (rent and area). For instance, the range 0–3.1 for 
larea corresponds to the range 0–22 ­m2 without any logarithmic 
transformation.
66805
Environmental Science and Pollution Research (2022) 29:66793–66807