Transcript slides

The Economic Value of Green Buildingｓ
Yongheng Deng
National University of Singapore
With
Junichiro Onishi
Chihiro Shimizu
Siqi Zheng
Xymax Real Estate Institute
Nihon University & NUS IRES
Tsinghua University
Hitotsubashi-RIETI International Workshop
on Real Estate Market, Productivity, and Prices
Tokyo, Japan, October 14, 2016
Introduction
• In the past decade, systems for rating and evaluating the
sustainability and energy efficiency of buildings have
proliferated.
•
•
•
•
•
•
•
14/10/2016
Energy Star and LEED (U.S.)
BOMA-Best (Canada)
BREEAM (UK)
HQE (France)
CASBEE (Japan)
Green Mark (Singapore)
Three Star System (China)
DENG @ Hitotsubashi-RIETI
2
Green Label Systems around the World
LEED
Energy Star
BREEAM
EPCs
HQE
GRESB
greenstar
Used in this study
-
-
-
-
-
-
-
Development
U.S.Green Building
Council (US)
U.S. Environmental
Protection Agency
（US）
Building Reserch
Establishment (UK)
UK Government
（UK）
HQE Association
(France)
GRESB
(Netherlands)
Green Building
Council of Australia
(Australia)
Since
1998
1995
1990
2006
1996
2010
2003
Target
Buildings
Buildings
Buildings
Buildings
Buildings
Firm
Buildings
Focus
Comprehensive
Energy Efficiency
Comprehensive
Energy Efficiency
Comprehensive
Comprehensive
Comprehensive
Output
4 ranks
Enegy Star ≧75
5 ranks
8 ranks
4 ranks
4 quadrants
6 ranks
Building and
equipment
performance
〇
-
〇
〇
〇
-
〇
Operation
〇
〇
〇
-
-
〇
〇
Water
〇
-
〇
-
〇
〇
〇
Material
〇
-
〇
-
〇
-
〇
〇
-
〇
-
〇
〇
〇
〇
-
〇
-
〇
-
〇
Transport
〇
-
〇
-
-
-
〇
Waste
〇
-
〇
-
〇
-
〇
Pollution
〇
-
〇
-
〇
-
〇
-
Manegement,
Performance
verification
Evaluation
items
Indoor
environment
Site and
Ecosystem
Other
14/10/2016
-
-
DENG @ Hitotsubashi-RIETI
-
Organization,
Disclosure, Risk
assessment, Green
lease
Management,
Innovation
3
Green Label Systems around the World
NABERS
CASBEE
CASBEE for real
estate
Used in this study
-
〇
〇
〇
〇
-
Development
Australian
Government
(Australia)
Ministry of Land,
Infrastructure,
Transport and
Tourism
（Japan）
Ministry of Land,
Infrastructure,
Transport and
Tourism
（Japan）
Development Bank of
Japan
(Japan)
Sumitomo Mitsui
Bank Corporation
（Japan）
Ministry of Land,
Infrastructure,
Transport and
Tourism
（Japan）
Since
1990
2004
2012
2011
2011
2014
Target
Buildings
Buildings
Buildings
Buildings
Buildings
Buildings
Focus
Energy Efficiency
Comprehensive
Comprehensive
Comprehensive
Comprehensive
Energy Efficiency
Output
5 ranks
5 ranks
4 ranks
5 ranks
6 ranks
5 ranks
Building and
equipment
performance
-
〇
〇
〇
〇
〇
Operation
〇
-
〇
-
〇
-
Water
〇
〇
〇
〇
〇
-
Material
-
〇
〇
-
〇
-
〇
〇
〇
〇
〇
-
-
〇
〇
〇
〇
-
Transport
-
〇
〇
〇
〇
-
Waste
〇
-
-
〇
〇
-
Pollution
-
〇
-
-
〇
-
-
Earthquake
resistance,
Handicapped
accessible
Earthquake
resistance, Useful
life, Disaster risk
Environment risk,
Crime prevention,
Tenant relation
Risk management,
Management policy,
Innovation
-
Evaluation
items
Indoor
environment
Site and
Ecosystem
Other
14/10/2016
DBJ Green Building SMBC Sustainable
Cetificate
Building Assessment
DENG @ Hitotsubashi-RIETI
BELS
4
Emerging Green Label System in China
14/10/2016
DENG @ Hitotsubashi-RIETI
5
Introduction
• Buildings account for approximately 40 percent of the consumption of
raw materials and energy.
• In addition, 55 percent of the wood that is not used for fuel is consumed
in construction.
• Overall, buildings and their associated construction activity account for
at least 30 percent of world greenhouse gas emissions.
• Consequently, energy represents the single largest and most manageable
operating expense in commercial building operations.
• UN Sustainable Development Goal (SDG) by 2030
• SDG 11 – Make cities and human settlements inclusive, safe, resilient and
sustainable
• SDG 13 – Take urgent action to combat climate change and its impacts
14/10/2016
DENG @ Hitotsubashi-RIETI
9
Eichholtz, Kok and Quigley (2010)
• Doing Well by Doing Good? Green Office Buildings (American
Economic Review, 2010)
• The seminal paper provides the first systematic analysis of the impact of
environmentally sustainable building practices upon economic outcomes as
measured in the marketplace.
• They find that buildings with a “green rating” command rental rates that are roughly
3 percent higher per square foot than otherwise identical buildings after controlling
for the quality and the specific location of office buildings.
• Premiums in effective rents are even higher, above 7 percent.
• Selling prices of green buildings are higher by about 16 percent.
• The percent increase in rent or value for a green building is systematically greater in
smaller or lower-cost regions or in less expensive parts of metropolitan areas.
• The private market does incorporate the “green” certification information in the
determination of rents and asset prices.
14/10/2016
DENG @ Hitotsubashi-RIETI
10
Deng, Li and Quigley (2012)
• Economic Returns to Energy-Efficient Investments in the Housing
Market: Evidence from Singapore (Regional Science and Urban
Economics, 2012)
•
•
•
•
•
•
14/10/2016
It provides one of the first analyses of the economics of green building in the residential
sector, and the first one analyzing property markets in Asia.
They adopt a two-stage research design
• In the first stage, a hedonic pricing model is estimated based on transactions involving
green and non-green residential units in 697 individual projects or estates;
• In the second stage, the fixed effects estimated for each project are regressed on the
location attributes of the projects, as well as control variables for a Green Mark rating.
Their results suggest that the economic returns to green building are substantial.
The returns vary by Green Mark category – both Platinum and Gold are positive and
statistically significant.
Green Mark Platinum remains significant using PSM nearest neighbor matching of control
and treatment samples.
The study provides insight about the operation of the housing market in one country, but the
policy implications about the economic returns to sustainable investments in the property
market may have broader applications for emerging markets in Asia.
DENG @ Hitotsubashi-RIETI
11
Zheng, Wu, Kahn and Deng (2012)
• The Nascent Market for “Green” Real Estate in Beijing (European
Economic Review, 2012)
• Using two unique geo-coded micro data sets to explore the nascent “green housing
market” in Beijing, from both the supply side and demand side.
• Based on Google searches, they construct a sample that contains information
whether certain housing complex’s greenness-related characteristics are emphasized
during its marketing.
• Focusing on information that developers wish to convey to potential buyers.
• The study found nascent “green housing market” does exist in Beijing.
•
•
On the Supply side – “Greenness” has been adopted as a marketing point in part of the
newly-built complexes, which helps gain significant price premium for their developers.
Although whether such “greenness” is really effective remains a question.
On the Demand Side – Pro-environmental households do have a smaller carbon footprint,
which provide potential demands for green housing.
• An introduction of a standardized official certification program would help ‘‘green’’
demanders to acquire units that they desire and would accelerate the advance of
hina’s nascent green real estate market.
14/10/2016
DENG @ Hitotsubashi-RIETI
12
Deng and Wu (2014)
• Economic Returns to Residential Green Building Investment: The
Developers’ Perspective (Regional Science and Urban Economics, 2014)
• The study provides the first evidence of the mismatch that developers face between
outlays and benefits in the residential green building sector. This mismatch may
impede further development of green residential properties.
•
•
•
The study found that the “green price premium” of residential developments arises
largely during the resale phase, relative to the presale stage. The market premium of
GM-rated units is about 10% at the resale stage, compared to about 4% during the presale
stage.
This implies that, while developers pay for almost all of the additional costs of energy
efficiency during construction, they only share part of the benefits associated with such
green investments.
The study found no evidence that the development of green housing can immediately and
significantly improve the corporate financial performance of Singaporean residential
developers.
• The emerging green real estate markets should be encouraged to introduce
innovative business arrangements and financial products that allow residential
developers to capture the future benefits associated with green properties.
14/10/2016
DENG @ Hitotsubashi-RIETI
13
Wu, Deng, Huang, Morck, and Yeung (2014)
• Incentives and Outcomes: China’s Environmental Policy (Capitalism
and Society, 2014)
1:45PM GMT 26 Feb 2013
“A city government’s spending on environmental improvements is actually
significantly negatively related to the odds of its (Communist party) secretary
and mayor being promoted,” wrote Professors Wu Jing, Deng Yongheng, Huang
Jun, Randall Morck and Bernard Yeung.
14/10/2016
DENG @ Hitotsubashi-RIETI
14
Deng, Onishi, Shimizu and Zheng (2016)
•
•
The Economic Value of Environmental Consideration in the Tokyo
Office Market (Working Paper, 2016)
We are among the first to analyze the economic value of green
buildings in an office market in Japan ;
• A social system incorporating environmental consideration has already come
into existence.
• Further environmental considerations would not necessarily lead to market
efficiency.
• Buildings constructed to high specifications in recent years are implicitly
meeting or exceeding environmental standards even without green labeling.
•
We use an independent data set, including 6,758 rental properties
in Tokyo office market and examine if there is a significant
premium from the green label for new contract rent.
14/10/2016
DENG @ Hitotsubashi-RIETI
15
Hedonic price function
The new contract rent of office buildings, based on
characteristics of office buildings and green label acquisition
conditions, is generally expressed as a hedonic price function.
𝑅𝑖 = h 𝑔𝑟𝑒𝑒𝑛𝑖 , 𝑥𝑖
(1)
𝑅𝑖 :
the new contract rent of completed contract case 𝑖
𝑔𝑟𝑒𝑒𝑛𝑖 : the green label dummy
𝑥𝑖 :
the vector that expresses the attributes in the completed contract case 𝑖
14/10/2016
DENG @ Hitotsubashi-RIETI
16
The model
Our specification is set up as following:
ln 𝑅𝑖 = α + 𝑔𝑟𝑒𝑒𝑛𝑖 ′ ∙ 𝛽 + 𝑥𝑖 ′ ∙ 𝛾 + 𝜀𝑖
ln𝑅𝑖 :
α:
β, γ:
ε:
𝑔𝑟𝑒𝑒𝑛𝑖 :
𝑥𝑖 :
14/10/2016
(2)
New contract rent logarithm (dependent variable)
Constant term
Vector of the coefficients correspondeing to each independent variable
Error term
Green label dummy (independent variable)
Vector that expresses the characteristic in the completed contract case 𝑖
(independent variable)
DENG @ Hitotsubashi-RIETI
17
Independent variables (1)
• Scale ;
• Gross building area, standard story area, and number of above-ground
stories.
• Age ;
• Age of the property and whether a renovation had been carried out.
• Convenience ;
• Distance to the nearest station and the presence or absence of office
specifications that consumers find appealing, such as a raised floor,
individual air conditioning, and automated security.
14/10/2016
DENG @ Hitotsubashi-RIETI
18
Independent variables (2)
• An office area dummy ;
• A dummy variable obtained by dividing a typical office area in
Tokyo’s 23 wards by 50 areas, in order to express the effects of
location on rent.
• In popular areas, rent tends to be high even for properties that are
medium scale, small scale, or old.
• The dummy for the time the contract was completed ;
• The demand-supply balance in the market at the time the property is
offered affects the rent.
• For quality adjustment at the time the contract was completed, the
two-year sample period from January 2013 to December 2014 was
divided into eight quarters and a quarterly dummy variable was
allocated according to the time the contract was completed.
14/10/2016
DENG @ Hitotsubashi-RIETI
19
Independent variables (3)
• Green label dummy
• If a building had acquired any one of the green labels used for the
analysis, the green label dummy was given a value of 1.
• CASBEE, CASBEE Real estate, DBJ Green Building Certification,
SMBC Sustainable Building Assessment
• There were three reasons for selecting these systems.
• Acquired on the unit of buildings
• Not the unit of corporations and portfolios
• Assessing comprehensive environmental performance
• Not only energy-saving performance
• A third party organization carries out the assessment based on the
standards that these systems have established.
14/10/2016
DENG @ Hitotsubashi-RIETI
20
Green labels used in this study
LEED
Energy Star
BREEAM
EPCs
GRESB
CASBEE
CASBEE for real
estate
DBJ Green
Building
Cetificate
SMBC
Sustainable
Building
Used in this study
-
-
-
-
-
〇
〇
〇
〇
Development
U.S.Green
Building Council
(US)
U.S.
Environmental
Protection
Agency
（US）
Development
Bank of Japan
(Japan)
Sumitomo Mitsui
Bank Corporation
（Japan）
Since
1998
1995
1990
2006
2010
2004
2012
2011
2011
Target
Buildings
Buildings
Buildings
Buildings
Firm
Buildings
Buildings
Buildings
Buildings
Focus
Comprehensive
Energy Efficiency
Comprehensive
Energy Efficiency
Comprehensive
Comprehensive
Comprehensive
Comprehensive
Comprehensive
Output
4 ranks
Enegy Star ≧75
5 ranks
8 ranks
4 quadrants
5 ranks
4 ranks
5 ranks
6 ranks
Building and
equipment
performance
〇
-
〇
〇
-
〇
〇
〇
〇
Operation
〇
〇
〇
-
〇
-
〇
-
〇
Water
〇
-
〇
-
〇
〇
〇
〇
〇
Material
〇
-
〇
-
-
〇
〇
-
〇
〇
-
〇
-
〇
〇
〇
〇
〇
〇
-
〇
-
-
〇
〇
〇
〇
Transport
〇
-
〇
-
-
〇
〇
〇
〇
Waste
〇
-
〇
-
-
-
-
〇
〇
Pollution
〇
-
〇
-
-
〇
-
-
〇
-
Manegement,
Performance
verification
Earthquake
resistance,
Handicapped
accessible
Earthquake
resistance, Useful
life, Disaster risk
Evaluation
items
Indoor
environment
Site and
Ecosystem
Other
14/10/2016
-
Building Reserch
UK Government
Establishment
（UK）
(UK)
-
GRESB
(Netherlands)
Organization,
Disclosure, Risk
assessment,
Green lease
DENG @ Hitotsubashi-RIETI
Ministry of Land, Ministry of Land,
Infrastructure,
Infrastructure,
Transport and
Transport and
Tourism
Tourism
（Japan）
（Japan）
Environment
Risk
risk, Crime
management,
prevention,
Management
Tenant relation policy, Innovation
21
Data
•
•
We constructed an integrated data on rent and green label
acquisition conditions.
Rental data ;
• Data used in this study are the contract case database for office buildings
collected by the Xymax Corporation.
• This dataset contains contract rent (not offered rent)
• 2013.1 – 2014.12
• Tokyo 23 wards
• 6,758 observations (2,689 buildings)
•
The characteristics of the buildings;
• Scale, Age, Facility, Location, Contract time
•
Green label dummy ;
• We collected and arranged the published information.
14/10/2016
DENG @ Hitotsubashi-RIETI
22
Variables
Variable
Unit
New contract rent
The contract rent when newly entering into the building (not the offered price when it
was being offered to tenants)
Green label dummy
Buildings granted a green label: 1
Gross building area
Gross building area of the building
sqm
Age
Number of years since construction
Year
Number of above-ground stories
Number of above-ground stories in the building
Standard story area
Standard story area of the building
sqm
Five city-center wards dummy
In the event that the building is located in one of the five city-center wards (Chiyoda
Ward, Chuo Ward, Minato Ward, Shinjuku Ward, and Shibuya Ward): 1 All others: 0
(0,1)
Time to nearest station
The number of minutes to walk to the building from the nearest station
Raised floor dummy
If a raised floor has been installed in the building: 1 All others: 0
(0,1)
Individual air conditioning dummy
If individual air conditioning has been installed in the building: 1 All others: 0
(0,1)
Automated security dummy
If automated security has been installed in the building: 1 All others: 0
(0,1)
Renewal dummy
Time of contract completion dummy
Area dummy
14/10/2016
Content
All others: 0
If the building has been/is being renewed at the time the contract was completed: 1 All
others: 0
On preparing dummy variables for each time of contract completion(quarter), If the time
the contract has been completed corresponds to the quarter : 1 All others ：0
On preparing dummy variables for each typical office area, in the event that there is a
building: 1 All others: 0
DENG @ Hitotsubashi-RIETI
Yen/sqm
(0,1)
Stories
M inutes
(0,1)
(0,1)
(0,1)
23
Summary Statistics
Number of Obsarvations = 6,758
mean
New contract rent (Yen/sqm)
standard
deviation
minimum
maximum
5,169.73
1,862.02
1,845.25
16,649.60
0.05
0.22
0.00
1.00
18,928.33
37,444.15
Age (Year)
23.73
11.83
0.00
59.91
Number of above-ground stories (Stories)
11.69
7.70
3.00
60.00
780.24
795.11
99.87
9,834.71
Five city-center wards dummy
0.77
0.42
0.00
1.00
Time to nearest station (Minutes)
3.36
2.31
0.00
15.00
Raised floor dummy
0.68
0.46
0.00
1.00
Individual air conditioning dummy
0.80
0.40
0.00
1.00
Automated security dummy
0.83
0.37
0.00
1.00
Renewal dummy
0.13
0.34
0.00
1.00
Green label dummy
Gross building area (sqm)
Standard story area (sqm)
14/10/2016
DENG @ Hitotsubashi-RIETI
992.56 379,447.90
24
Estimation results (baseline)
• The estimation result of ＋0.0439 was positive and
significant;
• The new contract rent for buildings that were granted a green label
was approximately 4% higher than for those that were not granted one.
Green label dummy
Number of samples
Percentage of buildings with green label
adjusted R-squared
0.0439***
(0.0115)
6,758
5.34%
0.6770
Notes: Standard errors in the parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01
• Whether acquiring a green label can result in higher returns?
14/10/2016
DENG @ Hitotsubashi-RIETI
25
Caveat
• In an economic analysis of environmental consideration, there
are concerns that factors such as scale and age of the property
become proxy variables for a green label.
• Scale:
• Many of the buildings that have been granted a green label were developed by
major developers and REIT; They are of large scale and are young buildings.
• They are able to pay the costs necessary to acquire a green label.
• Their shareholders and investors demand that they make such efforts.
• Age:
• The advance in environment and construction technologies for newly developed
properties
• If it is judged from the results of the preliminary survey that there is little
prospect of the building obtaining a high score and grade, few owners of old
office buildings would actually take the actions necessary to acquire certification.
14/10/2016
DENG @ Hitotsubashi-RIETI
26
Summary statistics : with/without green
label
• The average values for gross building area is 61,718 sqm and
age is 8.78 years in our green sample (properties with a green
label), showing that these buildings tend to be large scale and
newly developed.
14/10/2016
DENG @ Hitotsubashi-RIETI
27
Potential endogeneity
• When the decision to acquire the green label is also affected
by variables common to the newly rented office building,
such as size and age, endogeneity may occur in the estimation
method by identifying the new rent function using the green
label dummy.
• Therefore, when the hedonic function is estimated as a simple
linear model of the data in our analysis, the effects of size, age,
and performance reflect the difference in the presence or
absence of the green label, there is a potential bias in the
measurements.
14/10/2016
DENG @ Hitotsubashi-RIETI
28
Potential endogeneity
• We adopt the propensity score matching approach to make
sure the treatment group and control group have similar value
of the covariate.
• First; we estimate a variable that denotes the ease of which the green
label is obtained from the covariate (i.e., the propensity score).
• Next; based on the estimated propensity score, we create two similar
groups, and let the large difference between these groups indicate the
existence of a green label.
• Finally, we proceed with the analysis by estimating the effect of a
green label on a new contract rent.
14/10/2016
DENG @ Hitotsubashi-RIETI
29
Propensity score analysis (1)
• The effect of the green label on a new contract rent, when
treated as the difference between office buildings with or
without green label;
E 𝑌1 |𝐷 = 1 − E 𝑌0 |𝐷 = 0 .
𝑌1 :
𝑌0 :
D:1:
0:
14/10/2016
(3)
New contract rent, with green label granted
New contract rent, with green label not granted
When green label is granted to an office building
When green label is not granted to an office building
DENG @ Hitotsubashi-RIETI
30
Propensity score analysis (2)
• If the size or newness of a building affects whether there is
green label, it is possible that there is bias in the estimation
effect.
• We consider that the strict green label effect;
∆𝐷=1 𝒙 = E 𝑌1 − 𝑌0 |𝑃 𝒙 , 𝐷 = 1
(4)
𝒙: Characteristics of an office building that can be observed as a
covariate,
where note that 𝐱 is a vector of the components 𝑥1 , … , 𝑥𝑖 .
P 𝒙 = 𝑃𝑟 𝐷 = 1|𝒙 :
Forecasted probability of a green label being granted to an office building
with characteristics 𝐱.
14/10/2016
DENG @ Hitotsubashi-RIETI
31
Propensity score analysis (3)
• If we know what the new contract rent of an office building
with a green label was if it has not been granted, we can
simply extract the effect of the green label.
• Because it is normally not possible to observe such new
contract rents, the value that represents this is assigned a
weight by P 𝒙 , and the new rent is estimated from the new
rents of office buildings without green labels.
∆𝐷=1
𝑛1 :
14/10/2016
1
𝑥 =
𝑛1
𝑛1
𝑌1𝑖 𝒙𝒊 − 𝐸 𝑌0𝑖 |𝑃 𝒙𝒊 , 𝐷𝑖 = 0
𝑖=1
𝐷𝑖 =1
(5)
the number of samples that have been granted a green label
DENG @ Hitotsubashi-RIETI
32
Propensity score analysis (4)
•
New rents of office buildings that have not been granted a green
label, and are used as proxies, will have an expected value of;
𝑛0
𝐸 𝑌0𝑖 |𝑃 𝒙𝒊 , 𝐷𝑖 = 0 =
𝑊𝑖 𝑃 𝒙𝒊 𝑌0𝑗
𝑗=1
𝐷𝑗 =0
(6)
𝑛0 :
the number of samples that have not been granted a green label
W:
the weight assigned by P 𝒙 .
• In this study, nearest neighbor matching is used to assign weights for comparisons.
(Matching is performed on office buildings with the nearest probability of being
granted a green label.)
• The new contract rent of an office building that has not been granted a green label is
used as a proxy for new contract rent in the hypothetical case in which an office
building that has been granted a green label is an office building that has not been
granted a green label.
14/10/2016
DENG @ Hitotsubashi-RIETI
33
Propensity score analysis (5)
• Probit regression is used to estimate the probability of a green
label being granted to an office building possessing 𝒙
characteristics.
𝑥𝛽
1
𝑧2
P 𝒙 = Pr 𝐷 = 1|𝒙 = Φ 𝒙𝜷 =
𝑒𝑥𝑝 −
𝑑𝑧
2𝜋
2
−∞
β:
(7)
a vector of the elements 𝛽1 , … , 𝛽𝑖 ′.
• office characteristics for determining whether or not a green label is to
be granted:
• gross building area (sqm), standard story area (sqm), above-ground stories
(stories), age (years), area dummy, time to nearest station (minutes), raised floor
dummy, individual air conditioning dummy, automated security dummy, and
time of contract completion dummy.
14/10/2016
DENG @ Hitotsubashi-RIETI
34
Estimation results (Probit regression)
• The probability of acquiring a
green label is higher for high
quality office buildings that are
new, large, and equipped
with facilities.
Constant
Gross building area (logarithm)
-3.9439***
(0.5239)
0.4199***
(0.1331)
Age
Number of above-ground stories
-0.0727***
(0.0045)
-0.0146*
(0.0083)
Standard story area (logarithm)
-0.0492
(0.1522)
Time to nearest station
-0.0149
(0.0178)
Raised floor dummy
0.0938
(0.1368)
Individual air conditioning dummy
0.1596*
(0.0970)
Automated security dummy
0.1633
(0.1132)
Renewal dummy
1.0639***
(0.1257)
Time of contract completion dummy
Area dummy
Number of samples
Log likelihood
14/10/2016
Yes
Yes
6,758
-831.1474
AIC
1814.2950
Notes: Standard errors in the parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01
DENG @ Hitotsubashi-RIETI
35
Summary statistics (Propensity score
matching)
• By using propensity score matching, the differences between
the two groups were reduced, enabling us to create similar
samples.
Green-Label
Non-Green-Label
Number of Obserrvations = 361
Number of Obsarvations = 361
standard
deviation
minimum
maximum
7,638.37
2,154.42
3,025.00
16,649.60
7,625.18
2,283.37
2,420.00
13,612.50
Green label dummy
1.00
0.00
1.00
1.00
0.00
0.00
0.00
0.00
Gross building area
61,717.76
80,973.86
1,785.06
379,447.90
8.78
9.77
0.00
50.75
8.36
8.03
0.00
35.50
20.17
12.50
5.00
54.00
18.32
10.26
5.00
54.00
1,533.13
1,217.62
182.91
5,054.55
1,394.13
1,120.63
115.74
6,760.33
Five city-center wards dummy
0.86
0.35
0.00
1.00
0.84
0.37
0.00
1.00
Time to nearest station
2.98
1.80
0.00
10.00
2.81
2.16
0.00
10.00
Raised floor dummy
0.96
0.21
0.00
1.00
0.93
0.26
0.00
1.00
Individual air conditioning dummy
0.81
0.39
0.00
1.00
0.82
0.38
0.00
1.00
Automated security dummy
0.87
0.33
0.00
1.00
0.89
0.31
0.00
1.00
Renewal dummy
0.12
0.33
0.00
1.00
0.09
0.28
0.00
1.00
mean
New contract rent
Age
Number of above-ground stories
Standard story area
14/10/2016
DENG @ Hitotsubashi-RIETI
mean
standard
minimum maximum
deviation
48,572.06 56,058.70
1,100.83 263,996.70
36
Estimation results (Propensity score
matching)
• The estimated coefficient of the green label dummy was 1.6673 (1.8874) and, thus, was not a statistically significant
result, was negative, and close to zero.
• Green label does not have an effect on new contract rent, which
differs from Baseline.
Baseline
Green label dummy
Number of samples
Percentage of buildings with green label
adjusted R-squared
PSM
0.0439***
(0.0115)
-1.6673
(1.8874)
6,758
5.34%
0.6770
86
50.00%
0.9213
Notes: Standard errors in the parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01
14/10/2016
DENG @ Hitotsubashi-RIETI
37
Problems of propensity score matching
• The extracted samples by propensity score matching were
centered on large, newly constructed office buildings.
• The averages of gross building area and age for non-green buildings
were 48,572 sqm and 8.36 years.
• The data of the approximately 6000 non-green buildings that were not
matched were excluded.
• Green label has almost no effect, or even a slightly negative
effect on comparatively large and new office buildings.
However, the effect on other office buildings is unknown.
• In the analysis of the propensity score matching, it is
problematic that the effect of a green label on medium and
small office buildings and on old office buildings cannot be
confirmed.
14/10/2016
DENG @ Hitotsubashi-RIETI
38
Division of samples by clustering
• We divided the samples into five cluster based on the value of
the propensity score.
• Since the properties with a low propensity score tend to be smaller
and older, we should also be able to analyze the clusters of mid to low
propensity scores.
• Five quantile were used as the boundaries of these clusters, and the
number of samples in each cluster is nearly the same (number of
observations : 1351 or 1352 ).
14/10/2016
DENG @ Hitotsubashi-RIETI
39
Summary statistics (Cluster no.1: low
propensity score)
New contract rent
Green label dummy
Gross building area
Age
Number of above-ground stories
Standard story area
Five city-center wards dummy
Time to nearest station
Raised floor dummy
Individual air conditioning dummy
Automated security dummy
Renewal dummy
14/10/2016
Green-Label
Number of Obsarvations = 0
standard
minimum maximum
mean
deviation
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
NA
DENG @ Hitotsubashi-RIETI
Non-Green-Label
Number of Obsarvations = 1352
standard
minimum maximum
mean
deviation
9,831.25
4,131.48 1,211.07 1,905.75
0.00
0.00
0.00
0.00
995.04 190,590.00
9,714.12 21,009.43
58.75
0.00
8.76
25.50
60.00
3.00
4.98
9.41
9,666.12
102.48
653.62
620.70
1.00
0.00
0.50
0.44
13.00
0.00
2.20
3.55
1.00
0.00
0.49
0.58
1.00
0.00
0.39
0.82
1.00
0.00
0.39
0.81
1.00
0.00
0.27
0.08
40
Summary statistics (Cluster no.2:
medium low propensity score)
New contract rent
Green label dummy
Gross building area
Age
Number of above-ground stories
Standard story area
Five city-center wards dummy
Time to nearest station
Raised floor dummy
Individual air conditioning dummy
Automated security dummy
Renewal dummy
14/10/2016
Green-Label
Number of Obsarvations = 3
standard
mean
minimum maximum
deviation
6,453.33
462.08
6,050.00
6,957.50
1.00
0.00
1.00
1.00
9,579.06
5,332.62
3,421.49
12,657.85
44.39
10.87
31.83
50.75
8.33
1.15
7.00
9.00
653.31
225.10
393.39
783.27
1.00
0.00
1.00
1.00
4.00
1.73
3.00
6.00
0.67
0.58
0.00
1.00
1.00
0.00
1.00
1.00
1.00
0.00
1.00
1.00
0.67
0.58
0.00
1.00
DENG @ Hitotsubashi-RIETI
Non-Green-Label
Number of Obsarvations = 1348
standard
mean
minimum maximum
deviation
4,712.18 1,468.00 1,845.25 11,858.00
0.00
0.00
0.00
0.00
9,652.81 23,155.17
995.04 174,013.00
33.98
10.60
0.00
59.91
9.30
4.24
3.00
52.00
559.35
796.29
102.48
9,834.71
0.89
0.31
0.00
1.00
3.14
2.27
0.00
14.00
0.47
0.50
0.00
1.00
0.78
0.41
0.00
1.00
0.71
0.45
0.00
1.00
0.10
0.31
0.00
1.00
41
Summary statistics (Cluster no.3:
medium propensity score)
New contract rent
Green label dummy
Gross building area
Age
Number of above-ground stories
Standard story area
Five city-center wards dummy
Time to nearest station
Raised floor dummy
Individual air conditioning dummy
Automated security dummy
Renewal dummy
14/10/2016
Green-Label
Number of Obsarvations = 4
standard
mean
minimum maximum
deviation
4,840.00
1,864.74
3,630.00
7,562.50
1.00
0.00
1.00
1.00
14,906.88 22,039.92
1,995.57
47,904.50
25.79
4.46
20.16
31.08
12.25
7.89
7.00
24.00
482.22
322.34
244.63
958.45
0.50
0.58
0.00
1.00
3.50
1.91
1.00
5.00
0.75
0.50
0.00
1.00
1.00
0.00
1.00
1.00
1.00
0.00
1.00
1.00
0.00
0.00
0.00
0.00
DENG @ Hitotsubashi-RIETI
Non-Green-Label
Number of Obsarvations = 1348
standard
mean
minimum maximum
deviation
4,582.63 1,250.32 2,057.00 10,738.75
0.00
0.00
0.00
0.00
10,404.87 22,196.98 1,000.53 183,063.80
27.17
7.94
5.08
53.08
9.64
4.96
4.00
54.00
609.29
539.58
99.87
6,909.09
0.82
0.38
0.00
1.00
3.59
2.55
0.00
15.00
0.61
0.49
0.00
1.00
0.80
0.40
0.00
1.00
0.86
0.35
0.00
1.00
0.14
0.35
0.00
1.00
42
Summary statistics (Cluster no.4:
medium high propensity score)
New contract rent
Green label dummy
Gross building area
Age
Number of above-ground stories
Standard story area
Five city-center wards dummy
Time to nearest station
Raised floor dummy
Individual air conditioning dummy
Automated security dummy
Renewal dummy
14/10/2016
Green-Label
Number of Obsarvations = 30
standard
mean
minimum maximum
deviation
5,732.38
1,697.47
3,630.00
9,075.00
1.00
0.00
1.00
1.00
19,895.20 16,111.47
1,785.06
47,904.50
19.33
8.36
0.00
30.66
14.87
6.70
7.00
24.00
705.37
373.98
182.91
1,331.11
0.70
0.47
0.00
1.00
1.83
0.83
1.00
4.00
0.73
0.45
0.00
1.00
0.57
0.50
0.00
1.00
0.97
0.18
0.00
1.00
0.00
0.00
0.00
0.00
DENG @ Hitotsubashi-RIETI
Non-Green-Label
Number of Obsarvations = 1321
standard
mean
minimum maximum
deviation
5,201.66 1,420.37 2,420.00 11,495.00
0.00
0.00
0.00
0.00
14,673.38 19,420.47 1,036.36 162,406.60
22.19
7.88
0.00
54.08
11.34
5.94
3.00
50.00
760.92
582.13
105.79
5,983.47
0.84
0.36
0.00
1.00
3.53
2.31
0.00
14.00
0.81
0.39
0.00
1.00
0.80
0.40
0.00
1.00
0.89
0.31
0.00
1.00
0.21
0.41
0.00
1.00
43
Summary statistics (Cluster no.5: high
propensity score)
New contract rent
Green label dummy
Gross building area
Age
Number of above-ground stories
Standard story area
Five city-center wards dummy
Time to nearest station
Raised floor dummy
Individual air conditioning dummy
Automated security dummy
Renewal dummy
14/10/2016
Green-Label
Number of Obsarvations = 324
standard
mean
minimum maximum
deviation
7,860.37
2,093.80
3,025.00
16,649.60
1.00
0.00
1.00
1.00
66,650.90 83,908.30
2,992.86 379,447.90
7.26
8.44
0.00
33.00
20.87
12.82
5.00
54.00
1,630.89
1,242.78
320.00
5,054.55
0.87
0.33
0.00
1.00
3.07
1.84
0.00
10.00
0.98
0.14
0.00
1.00
0.83
0.38
0.00
1.00
0.86
0.35
0.00
1.00
0.13
0.34
0.00
1.00
DENG @ Hitotsubashi-RIETI
Non-Green-Label
Number of Obsarvations = 1028
standard
mean
minimum maximum
deviation
6,997.10 2,037.83 2,420.00 13,612.50
0.00
0.00
0.00
0.00
44,827.58 53,414.35
992.56 263,996.70
10.68
8.13
0.00
40.16
17.99
10.79
4.00
55.00
1,264.33
926.57
115.70
6,760.33
0.84
0.37
0.00
1.00
3.01
2.28
0.00
12.00
0.94
0.24
0.00
1.00
0.79
0.41
0.00
1.00
0.89
0.31
0.00
1.00
0.12
0.32
0.00
1.00
44
Estimation results (The hedonic function
for each cluster)
•
In cluster no. 5, (the large and new office buildings), the value is
negative and nearly zero (-0.0058 (0.0105)).
• the results from this cluster are consistent with those used propensity
score matching.
• In cluster no. 4 (the medium-sized and older office buildings), the result
was +0.1378 (0.0328), which is a significant positive effect.
• There were no statistically significant effects in cluster no.1 to 3.
• The effect of green label on new contract rents is not uniformly +4.3%
for all office buildings.
Baseline
Green label dummy
Number of samples
Percentage of buildings with green label
adjusted R-squared
Cluster
no.1
Cluster
no.2
Cluster
no.3
Cluster
no.4
Cluster
no.5
0.0439***
0.2453**
-0.0343
0.1378***
-0.0058
(0.0115)
(0.1231)
(0.0950)
(0.0328)
(0.0105)
1,351
0.22%
0.5034
1,352
0.30%
0.4852
1,351
2.22%
0.6218
1,352
23.96%
0.7684
6,758
5.34%
0.6770
1,352
0.00%
0.4669
Notes: Standard errors in the parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01
14/10/2016
DENG @ Hitotsubashi-RIETI
45
Robustness check
•
•
•
•
Office buildings that were granted green label had a higher propensity
score than office buildings that were not granted green label.
Although we have extracted the samples through clustering to those
with relatively closer propensity scores, there may still be a possibility
to remove the effect caused by the covariates from each cluster.
Therefore, we have chosen to conduct further propensity score matching
within the clustered sample groups, extract samples, and then estimate
the hedonic function.
We will be able to verify the effect of green label in samples with
greater homogeneity, which are not influenced by covariates.
14/10/2016
DENG @ Hitotsubashi-RIETI
46
Estimation results (Propensity score
matching within the clustered samples)
•
•
•
We extracted the samples and conducted further propensity score
matching within the cluster no.4.
The coefficient of the green label dummy was +0.1297 (0.0370), a
significant estimation result.
This result is equivalent, proving the robustness of our analysis.
PSM
Cluster
no.4
Green label dummy
Number of samples
0.1378***
0.1297***
(0.0328)
(0.0370)
1,351
Percentage of buildings with green label
adjusted R-squared
Cluster
no.4
2.22%
0.6218
60
50.00%
0.7964
Notes: Standard errors in the parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01
14/10/2016
DENG @ Hitotsubashi-RIETI
47
Extended model estimation
• In the probit regression, we learned that newer buildings are
more likely to be granted a green label.
• Even with clustered data, a certain relation will still exist
between the age of the building and the green label dummy.
• we make our estimate based on a model that adds the green
label dummy and the age as cross-terms to equation (2).
• Where, the age will be a true value instead of a dummy variable, since
there are only 30 cases in the data that have been granted green labels
within the samples in cluster no.4.
14/10/2016
DENG @ Hitotsubashi-RIETI
48
Estimation results (Include the age and
green labal dummy as cross-term)
•
•
•
In cluster no.4, the coefficient of the green label dummy was +0.3888
(0.0824), showing a significant positive result.
And the coefficient of the cross-term comprising the green label dummy
and the age of the building was -0.0131 (0.0039), showing a significant
negative result.
The Effect diminishes with time, and, finally, no further effect exists at
about 30 years from the original construction.
Cluster
no.4
Green label dummy
Cross term
(Green×Age)
Cluster
no.4
0.1378***
(0.0328)
Green label dummy × Age
Number of samples
Percentage of buildings with green label
adjusted R-squared
0.3888***
(0.0824)
-0.0131***
(0.0039)
1,351
1,351
2.22%
0.6218
2.22%
0.6247
Notes: Standard errors in the parenthesis. * p < 0.1, ** p < 0.05, *** p < 0.01
14/10/2016
DENG @ Hitotsubashi-RIETI
49
Conclusion
• After performing a quality adjustment of building
characteristics using the hedonic approach, green labels
showed a significant effect of +4.39% on new contract rents.
• By estimating the propensity score for the target variable of
the presence or absence of a green label, we confirmed that
the building characteristics of age and size make it easier to
obtain a green label.
• In order to address the problem of endogeneity, we used
samples that were adjusted using the nearest neighbor
matching technique, based on the propensity scores. In this
case, the effect of a green label was not statistically significant.
14/10/2016
DENG @ Hitotsubashi-RIETI
50
Conclusion
•
•
•
When we analyzed the samples clustered according to the value of
the propensity score, we found that the effect on the stratum in
which large and new buildings were concentrated was not
statistically significant, at -0.58%. Furthermore, the effect on the
stratum in which medium-sized and old buildings were
concentrated was +13.78%.
By conducting further propensity score matching on the clustered
samples, we were able to verify the robustness of our analysis
results.
There was a significant correlation between cross-terms regarding
the age of the building and the green label. We verified that, even
in the mid-size market, where green labels made a difference in the
contract rent, buildings that are 30 years or older are no longer
affected by these labels.
14/10/2016
DENG @ Hitotsubashi-RIETI
51