ArticlePDF Available

Green Buildings in Commercial Mortgage Backed Securities: The Effects of LEED and Energy Star Certification on Default Risk and Loan Terms


Abstract and Figures

We study the impact of green building on loans in the CMBS market. A hazard model shows green buildings carry 34 percent less default risk, all else equal. A matched-sample analysis gives similar results. We attribute the effect to a loan-to-value channel, where risk is lowered by a green price premium. The benefit comes at least partly from the level of green achievement, not only the label itself. Loans on buildings that were green at loan origination have slightly better terms than loans on non-green buildings. That difference is growing over time, but the effect is economically small compared to default risk. This article is protected by copyright. All rights reserved
Content may be subject to copyright.
2017 V00 0: pp. 1–36
DOI: 10.1111/1540-6229.12228
Green Buildings in Commercial
Mortgage-Backed Securities: The Effects
of LEED and Energy Star Certification on
Default Risk and Loan Terms
Xudong An* and Gary Pivo**
We study the impact of green building on loans in the CMBS market. A hazard
model shows green buildings carry 34% less default risk, all else equal. A
matched-sample analysis gives similar results. We attribute the effect to a
loan-to-value channel, where risk is lowered by a green price premium. The
benefit comes at least partly from the level of green achievement, not only the
label itself. Loans on buildings that were green at loan origination have slightly
better terms than loans on nongreen buildings. That difference is growing over
time, but the effect is economically small compared to default risk.
Over the past decade, researchers have consistently shown green buildings ex-
hibit certain “green premia,” including higher rents and prices, compared to
otherwise similar buildings (e.g., Miller, Spivey and Florance 2008, Eichholtz,
Kok and Quigley 2010, Pivo and Fisher 2010, Wiley, Benefield and Johnson
2010, Fuerst and MacAllister 2011, Deng, Li and Quigley 2012, Reichardt
2014, Devine and Kok 2015, Freybote, Sun and Yang 2015, Holtermans and
Kok 2017). While those studies reveal a great deal about the equity side
of green real estate, little research has been done on the debt side of the
market. We address that gap by comparing the default risk and loan terms
for green and nongreen buildings in the commercial mortgage-backed secu-
rities (CMBS) market. We find green buildings carry a large and previously
unreported benefit: lower default risk.
According to the U.S. Environmental Protection Agency (EPA 2016), green
buildings are “environmentally responsible and resource-efficient.” Various
*Federal Reserve Bank of Philadelphia, Philadelphia, PA and San Diego State Uni-
versity, San Diego, CA or
**University of Arizona, Tucson, AZ or
C2017 American Real Estate and Urban Economics Association
2An and Pivo
programs exist to certify them (Suzer 2015); however, we use “green build-
ings” to refer to those “registered” or “certified” under LEED and/or Energy
Star, the most common programs in the U.S. market.1Accordingly, “green,”
“green status,” and “being green” refer to being registered or certified under
one or both programs, while “green effect” refers to the impact of being green
on default risk or loan terms.
Empirically identifying the green effect is challenging due to a well-known
endogeneity issue: a building’s green status and financial performance are
subject to common drivers, which can be unobservable or difficult to control
in regressions. For example, green buildings are frequently in prime loca-
tions, which help their economic performance. The unsuccessful control of
unobservable drivers can lead to endogeneity bias.
To address this issue, we exploited both cross-sectional and time series vari-
ations in Energy Star and LEED status among CMBS buildings. As detailed
later, over 80% of the green buildings in our study were not green at loan orig-
ination; they became so later in the life of their loan. Therefore, to alleviate
the problem of unobservable loan and building characteristics, we compared
the default risk of the same loans before and after their collateral buildings
became green, which helped isolate the green effect. Meanwhile, the cross-
sectional variation in green status among buildings, as well as variation in
the timing of its appearance in our sample, helped identify effects associated
with time and observable loan and building characteristics.
In our analysis of loan terms, we again leveraged the cross-sectional and
time series variations in the timing of buildings becoming green. Obviously,
becoming green five years after loan origination would not impact original
loan terms. Therefore, in loan term regressions, we not only compared loans
on green and nongreen buildings, we also compared buildings that were
green at and after loan origination. While the effect associated with green
recognition after loan origination could be from unobservable differences
between green and nongreen buildings, any additional effect from green status
at loan origination should reflect the benefit of being green.
Our main data were provided by Trepp and the U.S. Green Building Council
(USGBC). All buildings in their data sets were geocoded so the data sets
could be merged based on the longitude and latitude of buildings. Among the
building types in our data, offices had the most green buildings, while other
1The distinction between “registered” and “certified” is important to our analysis, as
we explain later. We will also describe Energy Star and LEED in more detail, but
readers can also refer to Suzer (2015) or Vierra (2016) for more information.
Green Buildings in Commercial Mortgage-Backed Securities 3
types had very few. Therefore, our analysis was limited to CMBS loans for
office buildings.
Our hazard model showed being green led to a 34% reduction in loan default
risk, all else equal. To further alleviate concerns about uncontrolled locational
differences, we conducted a matched-sample analysis. For each green build-
ing, we found nearby buildings comparable in size, age, value per square
foot and local accessibility. Based on 223 green buildings matched to 373
nongreen buildings, we reran the hazard models and found the effect largely
remained, at 27%.
In our data, we not only had indicators of green status and timing, we ob-
served the LEED “points” or Energy Star “scores” earned during certification,
reflecting achievement on green metrics such as energy and water efficiency
(see “Background” for details). If the effect of being green on default is solely
due to being certified in and of itself, the points or score should not matter.
However, we found significant variation in default risk with respect to green
achievement. This finding also lessens the concern the green effect is from
uncontrolled locational difference.
We considered possible explanations for why green buildings carry lower
default risk. As noted, green buildings produce green rent premia, which
can be dynamic and hedge against risk in down markets (Das, Tidwell and
Ziobrowski 2011). These rent premia can increase the mortgage debt service
coverage ratio (DSCR), which lowers default risk. We found a small positive
correlation between DSCR and being green, but used contemporaneous DSCR
as a control variable, which absorbed the cash flow effects on default risk.
We also explored the idea that lower default risk is explained by equity
premia. As noted, studies show green value premia, consistent with their
higher income and lower risk. As we explain later, when most loans in our
study were being originated, bankers probably ignored or were unaware of
the green value premia. Moreover, most of the green buildings studied were
registered or certified years after loan origination. So, the green value premia
likely improved the equity position for owners and reduced the contempo-
raneous loan-to-value ratio (LTV), which is also known to lower default
On loan terms, we found in the matched-sample, green certification at loan
origination was associated with a 15- basis point (bps) interest rate reduction.
We also found the effect from certification grew stronger over time. However,
the effect on loan terms was financially insignificant compared to the effect on
default risk. These results are consistent with evidence presented that lenders
4An and Pivo
and appraisers largely ignored green features when most loans in the study
were underwritten.
Our study makes three contributions. First, we report direct evidence that
green buildings carry lower mortgage default risk and explore why this oc-
curs. The only published studies touching on this to date associate lower de-
fault risk with certain green residential attributes, including energy efficiency
(Kaza, Quercia and Tian 2014), transit service (Pivo 2013) and walkability
(Rauterkus, Thrall and Hangen 2010; Pivo 2013, 2014); but they do not test
the effect of being registered or certified under LEED or Energy Star.
Second, we address a particularly important question about loan pricing,
which is whether debts collateralized by green buildings enjoy lower cost.
Eichholtz et al. (2015) is the only study we find on that issue. It shows
green certified commercial real estate held by Real Estate Investment Trusts
(REITs) is financed at lower spreads, but it does not track the performance of
loans. In contrast, the match between ex ante loan pricing and ex post loan
performance in our study helps us better understand why there might be a
pricing effect.
Third, from a methodological perspective, studies on green real estate can be
susceptible to endogeneity due to unobservable factors, especially locational.
Our hazard model with time-varying green variables and our matched-sample
analysis mitigates that concern.
At a practical level, this study has implications for business. Results suggest
it can promote green building not only for its environmental merit, but also
for its economic benefit. Moreover, green building developers and investors
can argue for more liberal mortgage terms, while lenders and investors can
justify them as well.2From an investment perspective, the results can support
new CMBS products targeting green-labeled buildings.
In the next section, we present background information on green buildings
relevant to our analysis. We follow with descriptions of our data, methods
and results, and end with a summary and conclusions.
Energy Star is a voluntary program that allows owners and managers to
benchmark the energy performance of a building and earn the Energy Star
2Related to this, Fannie Mae announced on February 6, 2015 a reduction in mortgage
interest rates for green-certified multifamily buildings.
Green Buildings in Commercial Mortgage-Backed Securities 5
“label” if the building is an “energy-efficient top performer.” The building’s
operational characteristics and a year of verified utility bills are entered in
a benchmarking tool that “scores” the building from 0 to 100, indicating
how it is performing on energy use intensity relative to similar buildings
nationwide.3The building is “certified” and earns the Energy Star label if
it scores 75 or higher, indicating it is more efficient than 75% of otherwise
similar buildings.
The Energy Star label is good for one year. Performance must be maintained to
use the label from year to year. In our analysis, we assumed that even though
certifications are yearly, there could be a persistence effect in operations or
market response that lasts for at least one year beyond the life of the label.
We also tested other specifications including the effect lasting just a year or
being permanent.
LEED certification is another voluntary program.4To be LEED certified, a
building must be qualified under one of four versions of LEED.5Version 1.0
was launched in 1998, 2.0 in 2000, 3.0 in 2009 and 4.0 in 2014. Steps to certi-
fication include registration, application, review and certification. Registration
indicates certification is being pursued and is normally public information that
could affect market perception and indicate the building is performing well
in terms of energy efficiency or other metrics that could have financial ben-
efits. Therefore, we used the date of registration instead of certification to
identify LEED buildings. Because we only used LEED-certified buildings in
the study, every registered building was eventually certified.
To be certified, a building must satisfy prerequisites, such as minimum
achievement in energy and water efficiency, and earn enough “points” across
3The tool controls for operational characteristics including size, weather conditions,
number of occupants, number of computers and hours of operation per week. Energy
Use Intensity (EUI) is the building’s energy use per square foot per year using the
total gross floor area of the building. Source EUI is used to calculate the Energy Star
scores. It includes the raw fuel needed to operate the building including transmission,
delivery and production losses. Site EUI is the use reflected in energy bills and as
such is more relevant to the financial performance of buildings.
4Certain cities require LEED certification for government and private commer-
cial buildings. In Miami, San Francisco and San Jose, any commercial building
over 25,000 square feet must be LEED. In Seattle, Philadelphia or Indianapo-
lis private LEED projects receive incentives such as expedited permits or density
bonuses. See
5It is also possible for an entire neighborhood to become LEED certified, although
Freybote Sun and Yang (2015) found no effect from that on condos in their study of
Portland, Oregon.
6An and Pivo
seven topic areas.6A certification “rating level” is then awarded based on
the number of points earned. The rating levels are certified, silver, gold and
platinum, corresponding to higher levels of points earned and environmental
performance. About 86% of the LEED buildings studied earned a silver or
higher rating and more than half were gold or platinum.
All LEED versions since 2.0 included rating systems for specific situations,
such as LEED for New Construction (LEED-NC) and LEED for Existing
Buildings: Operations and Maintenance (LEED-EB). About 92% of the LEED
buildings in our sample were LEED-EB. Kok, Miller and Morris (2012) found
a 7% green value premium specific to LEED-EB buildings. The LEED-
EB system differs from LEED-NC and others by focusing on performance
during operations. As with Energy Star, historical performance records are
used to determine qualification; and certification must be renewed every five
years (U.S. Green Building Council 2014), which is reflected in the event-
histories for LEED-EB buildings in our hazard models. Certifications achieved
under LEED-NC and other rating systems are based on design standards and
Energy Star is exclusively concerned with energy efficiency, but energy effi-
ciency is also the largest point of emphasis in LEED. The most points that
can be earned from any single credit under LEED-NC 2009, for example, is
for optimizing energy performance (17%) and the most points that can be
awarded under any single topic area is for Energy and Atmosphere (32%).
The same is true for the 2008 and 2009 LEED-EB systems, which were used
to certify nearly all LEED buildings in our study. Moreover, both the 2008
and 2009 LEED-EB systems have a prerequisite that buildings earn an En-
ergy Star score of at least 69, or its equivalent; ensuring they are more energy
efficient than 69% of all similar buildings nationwide. While this emphasis
on energy efficiency was increased in the LEED 2009 program, it was also
the largest emphasis in prior versions (Wu et al. 2017).
Data and Methodology
Analytical Approach
As noted in the introduction, the endogeneity issue makes it challenging to
empirically identify green effects on default risk. For example, both green
6The areas are Sustainable Sites, Water Efficiency, Energy and Atmosphere, Materials
and Resources, Indoor Environmental Quality, Innovation in Design or Operations and
Regional Priority.
Green Buildings in Commercial Mortgage-Backed Securities 7
status and financial performance are subject to common correlates, prime
location among them.
To address this, we collected data that allowed us to exploit both cross-
sectional and time series variations in Energy Star and LEED status. About
86% of the green buildings in our sample were neither certified nor registered
at loan origination. Therefore, with longitudinal data and the hazard model we
could compare the default risk of the same loans before and after buildings
became green to help alleviate the problem of unobservable loan and building
characteristics, and thus isolate the green effect. Meanwhile, we used cross-
sectional variation in green status among buildings, as well as variation in the
timing of its appearance, to identify effects associated with time and other
observable loan and building characteristics.
Our main identification came from the hazard model using longitudinal data.
For each loan, we constructed quarterly event-histories from loan origination
to loan termination or the end date of data collection, whichever came first. In
the event-histories, we had the current green status of the loan, together with
other dynamic and static loan and building characteristics. The time-varying
green status variable captured the impacts of being green.
The hazard model was a standard Cox proportional hazard model, which is
widely used in the mortgage literature (see, e.g., Vandell et al. 1993, Seslen
and Wheaton 2010, An et al. 2013). In the model, the hazard rate of default
for a loan at a certain age/period is modeled as a function of a baseline hazard
function, which is a function of the duration (age) of the loan, and a function
of covariates, which are the default risk factors.
Assume the hazard rate of default of a mortgage loan at period Tas its
origination follows the form:
hi(T;Zi(t)) =h0(T)exp( Zi(t)β),i=1,...n.(1)
Here, h0(T) is the baseline hazard function, which only depends on the age
(duration), T, of the loan and is an arbitrary function that allows for a flexible
default pattern over time; Z(t) is a vector of covariates for an individual loan
that includes all the identifiable time-varying or time-invariant risk factors.
Our control variables included virtually all those identified by the existing
literature as significant drivers of commercial mortgage default risk or their
equivalents. More details on the variables are given later.
To further alleviate concerns about uncontrolled locational differences, we
used a matched-sample analysis. For each green office building, we found
8An and Pivo
nearby buildings comparable in size, age, value per square foot, Walk Score
and proximity to transit.
For the analysis of mortgage interest rate and other loan terms, we compared
buildings that were green at loan origination with those that became green
after loan origination, controlling for other differences in loan and building
features, to tease out the impact of green status on loan terms. The key
assumption here was green status achieved postorigination has no causal
impact on loan terms at origination.
CMBS Loan Data
Our CMBS loan data came from Trepp, Inc. Trepp partners with the Com-
mercial Real Estate Finance Council (CREFC) to gather detailed information
from monthly master servicer reports on all loans in the CMBS market. Master
servicer reports follow the CREFC Investor Reporting Package (IRP), which
provides for consistent data across all CMBS loans and collects information
from borrowers on the physical and financial status of their buildings.
We began with a Trepp loan data set covering nearly 90,000 loans, pooled
into 658 securities. We only requested data on loans collateralized by single
buildings, so they could be tied to a specific location for analytical purposes.
The data set included all CMBS loans in the Trepp database for office, re-
tail, multifamily and industrial buildings originating from 1998 through 2013.
We had monthly information on the status of each loan (prepaid, delinquent,
foreclosed or current) and the updated loan balance, DSCR, building occu-
pancy rate and loss information, if reported.7The monthly data on each loan
in the data set began when the loan was first originated and continued until
the end of 2013, or sooner if the loan was terminated earlier. The data set
also included static information on each loan including its origination date,
original balance, actual rate (mortgage note rate adjusted by points), maturity
term, amortization period, interest-only (IO) period, prepayment provisions,
originator, master and special servicers, securitization date (deal cutoff date),
face value, original LTV, net operating income (NOI) and DSCR at securiti-
zation. We also had information on the collateral property type, rentable area,
year built and location. The data set was comparable to that used by An et al.
7The CREFC IRP requires borrowers to provide regular updates of the current NOI,
occupancy rate and DSCR for each property. In some cases, servicers have been un-
willing to enforce such a rule, resulting in some missing values in updated occupancy
rates and DSCR in our data.
Green Buildings in Commercial Mortgage-Backed Securities 9
The loans analyzed were a portion of those originally obtained. We only used
fixed-rate loans and excluded adjustable rate mortgages, which comprised less
than 2% of the original data set. We only used CMBS loans from metropolitan
statistical areas (MSAs) covered by Real Capital Analytics (RCA) because
we needed RCA’s by-MSA and by-property type commercial real estate
price indices to calculate contemporaneous LTV. The RCA indices are only
available for some MSAs, albeit the largest ones in terms of CMBS loan
origination. We excluded loans originated before 2000 because that was when
RCA’s price indices began. We verified loan information on rate, LTV and
balance at origination and excluded a few loans with invalid information. A
representativeness check showed the remaining sample was representative of
the original CMBS database.
Location Data
As location plays an important role in real estate, we supplemented the Trepp
data by matching location information on each collateral property with lo-
cation metrics pertaining to walkability, which is akin to local accessibility,
and transit service. Both have been associated with lower mortgage default
risk and higher property values in commercial real estate (Pivo and Fisher
2011, Pivo 2013). We measured walkability using Walk Score provided by
Redfin. Walk Score rates the walkability of an address on a 100-point scale by
determining the distance to educational, retail, food, recreational and enter-
tainment destinations. Studies show Walk Score is a reliable, valid estimator
of neighborhood features linked to walking (Carr, Dunsiger and Marcus 2010,
2011, Duncan et al. 2011). We measured transit service using distance to the
nearest fixed rail transit station, based on station location maps published by
the Center for Neighborhood Technology.
Green Building Data
Our Energy Star and LEED certification information was provided by the
USGBC. They used building features, addresses and geocodes found in both
the USGBC and Trepp data sets to match loans in the Trepp data with
buildings in the USGBC green buildings database. For all matches identified,
they provided information on whether the matching buildings were LEED
and/or Energy Star certified, the date of certification (and registration for
LEED), the LEED version and system used for certification and the points or
score earned by the property during certification.
Our Office Sample
When we broke down our loan sample by building type, we found office
buildings had the highest percentage of Energy Star and LEED certifications,
10 An and Pivo
at about 11% and 4%, respectively. Less than 1% of the retail buildings
were Energy Star or LEED and even fewer certified multifamily or industrial
buildings.8Therefore, we limited our analysis to office buildings.
The final office loan sample used in our analyses contained 6,304 fixed-rate
mortgages originated from 2000 through 2013. The largest number (1,265)
was originated in 2006. A handful originated postcrisis in 2008, and none
in 2009. An increasing number originated after that. The collateral buildings
were located in the 17 MSAs with the largest office markets, including New
York, Los Angeles and Chicago.
The average original loan balance in our office sample was $24 million and
the average original LTV was 68%. The average original DSCR was 1.82 and
the average occupancy rate at origination was 95%. The average loan term
was 114 months (9.5 years) and the most popular loan term was 10 years.
The loans amortized much more slowly than they matured with an average
amortization term of 316 months (26 years), though the majority had a 30-year
amortization term. Consequently, many of the loans had balloon payments.
While all the loans in the final office loan sample were fixed rate mortgages,
a significant portion of them had IO terms. There were large variations in
the interest rates. The average was 6%, but they ranged from 4.7% to 15%.9
Table 1 contains descriptives for these and other variables in the office loan
sample such as building age, rental area, value per square foot and Walk Score.
Among the 6,304 buildings in our office loan sample, 673 (11%) were ever
Energy Star certified and 252 (4%) were ever LEED certified. By “were ever”
we mean they were certified during some or all of the timeframe covered by
our study (2000–2013). There are overlaps between Energy Star and LEED,
with 156 buildings (2.5%) having earned both.
Of the 252 LEED buildings, 233 (92%) were LEED-EB certified, which
means, as already mentioned, their certification was based on actual opera-
tional performance and required an Energy Star score of at least 69. Alto-
gether, 98% of our certified buildings were either Energy Star or LEED-EB,
meaning they were certified using records of actual operational performance
8There is no Energy Star certification program for whole multifamily buildings, only
individual units. There are also Energy Star and LEED programs for hotels. Zhang
et al. (2017) found a significant rate premium for green hotels in Beijing certified by
the China National Tourism Administration under their Evaluation Standard for Green
9As a robustness check, variables are winsorized at the 1st and 99th percentiles. All
results were consistent with our original findings.
Green Buildings in Commercial Mortgage-Backed Securities 11
Table 1 Summary statistics of securitized office loans.
Variable Mean Sum Std. Dev Minimum Median Maximum
Ever Energy Star 0.11 673 0.31 0 0 1
Ever LEED certified 0.04 238 0.20 0 0 1
Ever Energy Star or LEED certified 0.11 709 0.32 0 0 1
Ever Energy Star labeled and LEED certified 0.03 202 0.18 0 0 1
Original loan balance ($000s) 24,252 152,886,345 45,204 1,000 9,568 806,000
Interest rate (%) 6.00 37,796 0.88 4.66 5.81 15.00
Loan-to-value ratio (LTV, %) 67.75 427,116 12.22 23.40 71.30 92.33
Debt-service coverage ratio (DSCR) 1.82 3,048 0.82 1.17 1.60 3.90
Occupancy rate (%) 94.70 576,827 7.00 21.22 97.60 100
Term (months) 114 720,221 23 35 120 360
Amortization term (months) 316 1,992,570 111 35 360 432
Interest-only (IO) periods (months) 32 200,745 42 0 1 219
Weighted average mortgage constant 6.87 36,773 1.10 3.79 6.79 24.54
Age of the building 36 220,037 24 2 30 113
Property rentable area (sq. ft.) 184,743 1,164,250,451 311,263 8,349 81,793 3,781,045
Value per sq. ft. ($) 230 1,452,185 433 29 186 742
Public transit within 1/4 mile of the building 0.22 1,403 0.42 0 0 1
Property walk score (0100) 65 352,259 27 0 65 100
Notes: (1) Energy Star and LEED certification data from the U.S. Green Building Council; (2) Property walk score is obtained from Redfin
and matched to the CMBS loan based on the longitude and latitude of the building. It is used in our sample match later in the analysis;
(3) the weighted average mortgage constant is defined as the weighted average annual debt service per $100 loan balance using IO period
as a percentage of loan term as weight; (4) the office loan sample is from the original sample of nearly 90,000 CMBS loans. We focus on
fixed-rate loans covered by RCA price indices (in 17 MSAs and originated in 2000 or after). We lost a small portion of the loans during
geocoding and data match.
12 An and Pivo
showing they were more energy efficient than at least 69% of otherwise sim-
ilar buildings in the nation. This supported our expectation that the LEED
buildings enjoyed a green income premium from greater energy efficiency
and lower energy bills.
As noted, most (86%) of the buildings in our sample were green registered
or certified years after their loans were originated. Only 113 (1.8%) of the
office buildings were Energy Star certified at loan origination, 7 were LEED
certified and 6 were LEED registered. All others were LEED registered an
average of 4.9 years later, LEED certified an average of 6.3 years later and/or
Energy Star certified an average of 5.2 years later.
As noted above, LEED-EB certifications are valid for five years and Energy
Star for one year. One building had a LEED-EB certification that expired
14 months before the end of the study period. It was recorded as LEED-EB
through the end of the study in its event-history to account for the persistence
effect already discussed. Another two LEED-EB buildings expired before the
end of the study but were recertified with no gap in certification. All other
LEED-EB buildings had certifications that expired after the end of the study.
The Energy Star buildings in our office sample were certified for anywhere
from 1 to 14 years during the study period, with a mean of 3.4 years; however,
the certification years were not always contiguous. About 25% of the Energy
Star buildings had gap years (mostly one or two years), between their first and
last year of certification, when they were not certified. The frequency with
which we saw one-year gaps during several years of contiguous certifications
is one reason we recorded buildings as Energy Star in the event-histories for
the year they were certified and the subsequent year. As already noted, we
felt reputational or performance effects from certification probably persisted
during the gap years.
Initial Comparisons
Following the literature, we defined default as a 60-day delinquency in loan
payments. Among the 6,304 office loans in our sample, 1,001 were in default
at some point during the study period, producing an overall default rate of
15.9%. The default rate for buildings that were ever green, however, was
significantly lower.
Table 2 presents a simple comparison of the default rates for the certified
and noncertified office buildings. Those that were ever Energy Star certified
Green Buildings in Commercial Mortgage-Backed Securities 13
Table 2 Default rate of securitized office loans: green versus nongreen loans.
Energy Star
Energy Star or
LEED Certified
Both Energy
Star and LEED
0 16.7 16.3 16.7 16.2
1 9.4 5.9 9.5 5.0
Difference 43% 64% 43% 69%
Notes: (1) Default rate numbers in percent (%); (2) The sample includes 6,304 office
CMBS loans, among which 1,001 defaulted; (3) 673 buildings are ever-Energy Star
and 238 are ever-LEED certified. Among the 709 buildings that are Energy Star or
LEED certified, 202 buildings are both Energy Star and LEED and the rest have one
of the two green certifications. (4) Default is defined as 60-day delinquency.
had a default rate of 9.4%, compared to 16.7% for those that were not. In
other words, they had a 43% lower default rate than all other buildings. The
difference between were ever LEED certified, and all other buildings, was
even larger. Only 5.9% of LEED-certified buildings defaulted, compared to
16.3% for all other buildings, giving them a 64% lower default rate. Finally,
buildings with both certifications had the lowest default rate at 5.0%, or 69%
lower than those without.
Certainly, some of these differences could be due to differences in the loan
and building characteristics of the certified and noncertified groups that do
not pertain to being green. For example, certified buildings on average are
more expensive per square foot, possibly due to being in better locations or
higher class (Table 3). To account for such differences in loan and building
characteristics, we ran a simple Logit model where our dependent variable
was a default indicator (0 or 1) and our covariates included green dummies,
log loan balance, LTV at loan origination, occupancy rate, amortization term,
maturity term, property value per square foot, age of the building, as well as
MSA- and vintage-dummies. Note the value per square foot variable should
Table 3 Comparison of key variables for certified and noncertified loans/buildings.
LTV (%) Occupancy (%)
Value per sq.
ft. ($)
Noncertified 68% 95 221 16 36
Certified 64% 93 308 17 40
Note: These are respective sample means at loan origination.
14 An and Pivo
Table 4 Default rate differences for loans on certified/noncertified buildings after
controls: logit model results.
Modeled Difference in Default Rate
Model 1 Model 2 Model 3 Model 4
Energy Star 0.423***
LEED 0.773***
Energy Star or LEED 0.417**
Both Energy Star and LEED 0.929***
Control variables Loan characteristics, building characteristics, nearby
public transit, MSA-fixed effect and vintage-fixed
Model pseudo R-square 0.159 0.159 0.159 0.160
Notes: (1) Results from Logit models where the dependent variable is default or
not during the life of the loan (up to the data collection point); (2) The list of
control variables includes dummy for public transit within 1/4 mile of the building log
loan balance, origination LTV, origination occupancy rate, amortization term, maturity
term, property value per sq. ft., age of the building, MSA-fixed effect and vintage-fixed
effect; (3) ** for p <0.05 and *** for p <0.01.
serve as a strong control for many factors related to value such as location,
class and amenities.
The green certification dummies were defined as buildings that were ever
LEED or Energy Star certified. A caveat here is this variable could capture
some of the generic difference between green and nongreen buildings, not
necessarily the green effect per se. We address this issue in our hazard model.
Table 4 presents our Logit model results. It shows that after controlling
for loan and building characteristics, buildings that were ever Energy Star
buildings had a 35% lower default rate than other buildings, and buildings
that were ever LEED had a 54% lower default rate than others. Those that
were ever both Energy Star and LEED had the lowest default rate of all;
nearly 61% lower than for noncertified buildings.
Hazard Model Results
As explained earlier, our hazard model exploited both the cross-sectional
and time series variations in certification as well as the timing of becoming
Green Buildings in Commercial Mortgage-Backed Securities 15
certified in a hazard model to help our identification. The idea was since
only 1.8% of the Energy Star buildings were certified, and only about 3% of
the LEED buildings were registered or certified at loan origination, we could
compare the default risk of the same loan before and after its collateral prop-
erty becomes green. Such a comparison reduced the impact of unobservable
loan and building characteristics, and helped us identify the green effects on
default risk.
The hazard model also allowed us to account for the fact that Energy Star
certifications only last for one year and LEED-EB certifications for five years.
In the quarterly time series for the hazard model, a building was recorded
as LEED from the date it was registered in all subsequent quarters, except
for LEED-EB buildings where the designation was discontinued five years
after the certification date. Meanwhile, to account for the possible persistence
effects already discussed, a building was defined as Energy Star in the time
series for two years from the date it was certified. As previously noted,
we tried other definitions for Energy Star including a “temporary effect”
definition where we assume the effect lasts just one year from the registration
date and a “permanent effect” definition where we assumed the effect lasts
for the whole life of the loan. The impacts on the results were small and
reported in Appendix Table A3.
Our control variables in the hazard model included contemporaneous LTV
and DSCR, original loan balance (in log terms), original LTV, a dummy for
original LTV higher than 70%,10 refinance incentive (measured by percentage
decline in market prevailing mortgage interest rate relative to the current note
rate), building age (dummies for buildings less than 11 and more than 39
years old, which along with size, helped us address the quality endogeneity
issue and proxied for building class (as discussed by Pivo and Fisher 2010),
prepayment restrictions (i.e., the presence of a prepayment lock out, prepay-
ment penalty and yield maintenance in a particular loan quarter, which tend
to limit refinancing and increase default risk), MSA unemployment rate in-
novation in a particular quarter (i.e., the change in the MSA unemployment
rate over the prior 4-quarter moving average, as a business cycle indicator),
volatility of the 10-year Treasury rate, volatility of the RCA price index for
each of the 17 MSAs and building types over the prior 12 quarters and MSA-
and vintage-dummies.11 In addition, we included per square foot value of
10This follows the approach used by Titman, Tompaidis and Tsyplakov (2005) and
Ambrose, Shafer and Yildirim 2016 for addressing the endogeneity of mortgage
spreads and LTV, first discussed by Archer et al. (2002). We use the same approach
in our loan term regressions.
11For details on these control variables, please see An, Deng and Gabriel (2016) and
An, Deng, Nichols and Sanders (2013).
16 An and Pivo
the building as an additional control in order to capture most location and
building amenity-related factors.
To construct the contemporaneous LTV, we utilized the RCA price index (by
building type and by MSA) to bring building value up-to-date, and calcu-
late contemporaneous LTV as the ratio between current building value and
remaining loan balance in the Trepp data.
For contemporaneous DSCR and occupancy, if there was a quarter the servicer
did not report the current DSCR or occupancy rate we used the nearest
quarter’s value either before or after the missing quarter, as a proxy.
Table 5 compares descriptive statistics for the time-varying covariates for
the defaulted and nondefaulted loans at loan origination and termination.
Each observation was one loan in a specific quarter. It shows the mean
LTV at loan origination for defaulted loans was only about 10% higher
than for nondefaulted loans (70% vs. 64%), while at loan termination it was
almost 60% higher (93% vs. 59%). Not surprisingly, contemporaneous LTV is
shown later to be an important driver of default. The green dummy variables
demonstrate similar patterns. For example, there was little difference between
the percentage of defaulted and non-defaulted loans that are Energy Star at
loan origination, but at loan termination, about 10% of non-defaulted loans
were Energy Star compared to only 4% of defaulted loans.
Our main hazard model results are given in Table 6. We only present our
focus variables in the table and refer readers to Appendix Table A1 for
the full results. The statistics given are the regression coefficient, signifi-
cance and hazard ratio (in parenthesis). The hazard ratio is the predicted
change in default risk produced by a one-unit change in the parameter, hold-
ing other variables constant. A hazard ratio greater than 1 indicates default
risk increases as the variable increases and a hazard ratio less than 1 in-
dicates it decreases as the variable increases. For binary covariates (e.g.,
Energy Star), the hazard ratio estimates the ratio of the risk of default for
loans with and without the feature. For the continuous covariates (e.g., value
per square foot), because they are standardized, the hazard ratio estimates
the change in risk associated with a one standard deviation change in the
We have two model specifications: in model 1, Energy Star certification is the
green indicator and focus variable; in model 2, the focus variable is 1 if the
12In our model estimation, all continuous variables are standardized to zero mean and
unit variance.
Green Buildings in Commercial Mortgage-Backed Securities 17
Table 5 Descriptive statistics of time-varying covariates.
At Origination At Termination
Nondefault Default Nondefault Default
Variable Mean STD Med. Mean STD Med. Mean STD Med. Mean STD Med.
Energy Star 0.01 0.12 0 0.02 0.13 0 0.10 0.27 0 0.03 0.17 0
LEED 0.00 0.04 0 0.00 0.04 0 0.04 0.19 0 0.03 0.05 0
Energy Star or LEED 0.02 0.12 0 0.02 0.13 0 0.09 0.28 0 0.03 0.18 0
Energy Star and LEED 0.00 0.04 0 0.00 0.00 0 0.03 0.17 0 0.00 0.00 0
Contemporaneous LTV (%) 64.32 12.72 67.27 70.42 13.57 70.96 58.69 21.10 56.31 93.45 32.32 90.52
Current DSCR 1.72 0.55 1.58 1.56 0.34 1.53 1.65 0.61 1.54 1.25 0.43 1.25
Refinance incentive 0.01 0.08 0.01 0.02 0.10 0.03 0.25 0.10 0.27 0.26 0.12 0.25
MSA unemployment rate innovation 0.99 0.13 0.96 1.03 0.19 0.97 0.92 0.06 0.90 1.09 0.21 1.00
Vol. of 10-year Treasury rate 0.36 0.12 0.32 0.36 0.12 0.32 0.38 0.04 0.39 0.48 0.12 0.45
Vol. of CPPI 7.65 3.86 7.26 8.21 4.79 7.19 9.30 4.13 10.14 11.32 7.77 9.26
Number of observations 5,303 1,001 5,303 1,001
Notes: (1) Current LTV is calculated based on origination LTV and MSA-level commercial property price index (CPPI) from Real Capital
Analytics (RCA); (2) current DSCR is from the operating statement of the property; in the case of missing DSCR in the operating statement,
we use the nearby quarter result; (3) refinance incentive is calculated as the difference between the origination mortgage interest rate and
the current prevailing mortgage interest rate; (4) MSA unemployment rate innovation is the ratio between the current quarter unemployment
rate and past 6 quarter moving average; the variable is a used as a business cycle indicator (see Korniotis and Kumar 2013).
18 An and Pivo
Table 6 Impact of green label and certificate: default hazard model results.
Full Sample Matched Sample
Model 1 Model 2 Model 1 Model 2
Energy Star 0.411** 0.342**
(0.663) (0.710)
Energy Star or LEED 0.412** 0.314**
(0.662) (0.731)
Control variables Contemporaneous LTV, current DSCR, refinance
incentive, macroeconomic variables, loan
characteristics, building characteristics, nearby
public transit, MSA-fixed effect, vintage-fixed
effect and baseline hazard
Number of loans 6,304 596
Number of loan-quarters 183,425 14,105
Notes: (1) Hazard ratios are in parenthesis; (2) complete model results in the Appendix;
(3) the focus variable “Energy Star” and “Energy Star or LEED” are time-varying
variables as a building may get certified at a certain age of the loan after origination.
In the full sample, among the 6,304 loans, 673 are ever Energy Star labeled and
709 are ever Energy Star or LEED certified. Also, among those loans, 95 loans were
Energy Star at loan origination and 100 were either Energy Star or LEED certified at
loan origination; and at loan termination (censoring), 459 loans were Energy Star and
504 loans were either Energy Star or LEED certified. (4) In the matched sample, 223
green loans are matched to 373 nongreen loans. The match is based on building type,
ZIP code, property value per square footage, building age, walk score and building
Euclidian distance; (5) *for p<0.1, ** for p<0.05 and *** for p<0.01.
building is Energy Star (in the current or previous quarter) or LEED registered
(in any previous quarter). We did not use LEED separately because a very
small portion of the buildings were LEED registered or certified (especially
in the event-history sample where the percentage of quarters with a LEED
building is extremely low). Remember our focus variable is time-varying; for
the same loan the LEED indicator is zero before the building is registered
and 1 afterward.
In the full model results, we see Energy Star certification had a negative
impact on the default hazard rate and the impact was statistically significant
(model 1). From the hazard ratio, we see the impact was also economically
significant, as the Energy Star label reduced the hazard rate by 34% (1–0.663),
all else equal. In model 2, the result was the same: Energy Star certification or
LEED registration reduced the hazard rate by 34% (1–0.662), all else equal.
From Appendix Table A1, we can see our control variable results were highly
consistent with the existing literature. For example, contemporaneous LTV
Green Buildings in Commercial Mortgage-Backed Securities 19
was highly significant and positively related to default probability.13 Contem-
poraneous DSCR was also significant and negatively related to default proba-
bility.14 MSA unemployment rate innovation, a business cycle indicator, was
positive and significant, meaning when the local economy was weaker, the
chance of a loan defaulting was higher. Loans with high LTV (above 70%) at
origination and larger loans also showed higher probability of default, all else
equal.15 When there was lock out, the refinance incentive was significant and
positively related to default probability, which was also consistent with the
literature (see, e.g.,Anet al. 2013). Value per square foot was negative and
significant, meaning buildings in better locations or of better quality were less
likely to default. We also ran the same model with additional higher order
terms for LTV and DSCR (see Appendix Table A4), producing little change
in the results.
Matched-Sample Results
In addition to using the longitudinal data and hazard model to help identifica-
tion, we also conducted a matched-sample analysis. For each certified office
building, we first found office buildings in the same ZIP code of comparable
size, age, value per square foot, Walk Score and walking distance to transit.
For value per square foot and Walk Score, we only allowed a ±10% discrep-
ancy for a match. For age, we allowed a five years difference.16 For size,
measured by rentable area, we allowed a 50% difference.
13As An et al. (2013) pointed out, negative equity is a well-established reason for
mortgage default, making the contemporaneous LTV a natural risk factor. Several
studies document the positive relationship between contemporaneous LTV and default
risk (Vandell 1992, Episcopos, Pericli and Hu 1998, Goldberg and Capone 1998, 2002,
An et al. 2013, Chen and Deng 2013).
14Jones and Sirmans (2016) review six studies that show a negative relationship
between default risk and contemporaneous DSCR (Ciochetti et al. 2002, Goldberg
and Capone 2002, Ciochetti et al. 2003, Grovenstein et al. 2005, Seslen and Wheaton
2010, Cho, Ciochetti and Shilling 2013; also see An et al. 2013) and six that show a
negative relationship with original DSCR (Archer et al. 2002, Grovenstein et al. 2005,
Yildirim 2008, Cho, Ciochetti and Shilling 2013, Furfine 2014, Seagraves and Wiley
15Despite the observation by Archer et al. (2002) that it can be difficult to observe an
empirical relationship between the original LTV and default because bankers mitigate
risk by reducing LTV ratios at origination, several studies (reviewed by Jones and
Sirmans 2016) find there is a positive relationship between original LTV and default
(Yildirim 2008, Titman and Tsyplakov 2010, Furfine 2014, Seagraves and Wiley
16Five years was selected based on prior research showing that in 1996 the period of
greatest depreciation was 7–12 years and declining (Baum and McElhinney 2000).
Tenant interest in green building is also evolving, which could further change how
tenants view the effect of age on utility and value (Mansfield and Pinder 2008).
20 An and Pivo
After finding matches for each Energy Star-certified and LEED-registered
building, we rank ordered them based on the Euclidian distance between
the subject building and each match. Distance was calculated based on the
longitude and latitude of the two buildings. For green buildings with multiple
noncertified building matches, we selected the two closest matches. As not all
green buildings had matches, our final matched-sample contained 395 green
buildings matched to 609 nongreen buildings.
This matching algorithm mimicked the behavior of tenants and investors
searching for substitutes in the real estate market. Therefore, the matched-
sample analysis helped control unobservable locational differences.
We reran the hazard models using the matched sample.17 Our main results are
given in Table 6. There we see the focus variable results are weaker than with
the full sample but still economically significant. Energy Star certification
reduced the hazard rate by 29% (1–0.710), while being Energy Star certified
or LEED registered reduced the hazard rate by 27% (1–0.731).
The Mechanism
To this point, our analysis showed strong evidence that being green reduces
default risk in CMBS office loans. But we wanted to understand why.
For commercial mortgage loans, NOI is used to service the loan, and as green
buildings can be expected to have enhanced NOI, there could be a cash flow
or NOI channel through which green development reduces default risk.
NOI is the net result of rent, occupancy and operating cost. Many studies
show green buildings have higher rent and occupancy rates (Miller, Spivey and
Florance 2008, Eichholtz, Kok and Quigley 2010, 2013, Pivo and Fisher 2010,
Wiley, Benefield and Johnson 2010, Fuerst and McAllister 2011, Reichardt
et al. 2012, Das and Wiley 2014, Reichardt 2014, Devine and Kok 2015,
Robinson and McAllister 2015, Szumilo and Fuerst 2015, Robinson et al.
2016, Holtermans and Kok 2017). Moreover, the rent premia may be dynamic
and provide a hedge in down markets (Das, Tidwell and Ziobrowski 2011).
Meanwhile, studies on operating cost are mixed. One would expect lower
operating cost because LEED and Energy Star emphasize energy efficiency
and certified buildings are more energy efficient (Newsham, Mancini and Birt
2009, Scofield 2013, Devine and Kok 2015). But while Miller, Pogue and
Saville (2010) found lower energy bills in Energy Star buildings, they did
17The number of matches is used as a weight for the subject (green) buildings.
Green Buildings in Commercial Mortgage-Backed Securities 21
not find lower total operating cost, suggesting there may be other unobserved
expenses. Similar findings on Energy Star buildings were reported by Pivo and
Fisher (2010) and Reichardt (2014), though Reichardt found lower operating
cost in LEED buildings. Thus, it seems reasonable to expect higher NOI in
green buildings, though more likely due to higher rents and occupancy than
lower total expenses.
Higher NOI should lead to higher DSCR, which should reduce default risk.18
However, recall in our hazard model, we used contemporaneous DSCR as a
control variable, which absorbed the NOI effect. Therefore, one can interpret
the reductions in default found in that model as additional effects over and
above any benefits from green status to cash flow or NOI. A caveat here
is this interpretation assumes the green and DSCR variables are orthogonal.
In that regard, we first tested the correlation between the green certification
and DSCR. Results showed the correlation between Energy Star certification
and contemporaneous DSCR was only about 8%. Second, in an alternative
specification we included higher-order effects of DSCR in the model. We
found the green effect remained strong (see Appendix Table A4). Therefore,
we concluded the green effect we observed was most likely over and above
the impact of green status on the DSCR of the loan.
Lower default risk from LEED and Energy Star could also come through
an improved equity position or LTV channel. As discussed earlier, LTV
is a strong driver of mortgage default and studies show value premiums
in green buildings (see, e.g., Miller, Spivey and Florance 2008, Eichholtz,
Kok and Quigley 2010, 2013, Pivo and Fisher 2010, Wiley, Benefield and
Johnson 2010, Fuerst and McAllister 2011, Das and Wiley 2014, Robinson
and McAllister 2015). Those premiums could be either from the higher NOI
discussed above or from lower capitalization rates.19 Note in our hazard
model, our contemporaneous LTV control variable is calculated based on a
commercial building price index. Therefore, we think to the extent individual
green buildings had property value appreciation higher than normal buildings
in an MSA, any effect from higher appreciation on the equity position and
18This assumes the higher NOI was not forecasted ex ante by the lender who responded
with a larger loan, resulting in more debt service and a normal DSCR. However, as
indicated and discussed further under Loan Terms, we do not think lenders historically
recognized green premia in the lending process. Moreover, most green buildings in
our data set were certified well after their loans were originated, which would have
made impossible any response by lenders to expected premia.
19Studies confirm lower cap rates in certified buildings. Miller, Spivey and Florance
(2008) found them to be 55 bps lower for LEED buildings, Pivo and Fisher (2010)
found them to be 52 bps lower for Energy Star buildings, and McGrath (2013) found
them to be 36 bps lower for a combined sample of LEED and Energy Star buildings.
22 An and Pivo
LTV of the property, and thus the default risk of the loan, reflected the green
status of the building.
A related issue is whether the reduced default risk in green buildings is from
green certification and labeling by itself, or whether it changes with higher
levels of green performance, as reflected in the points or score earned in
the certification processes. Several prior studies show a positive relationship
between higher green building scores or points and rental rates (Dermisi 2009,
Fuerst and McAllister 2009, Eichholtz, Kok and Quigley 2010, 2013, Wiley,
Benefield and Johnson 2010, Fuerst and McAllister 2011, Szumilo and Fuerst
2015, Holtermans and Kok 2017), so it seemed reasonable to expect default
risk to decline with increasing green achievement.
To explore this issue, we devised another test. In our green building data,
we not only observed any green label and the date it was obtained, but also
green points received by each building during the certification process. If the
green effect purely arises from the green label itself, we should not have seen
differences in default risk among green buildings with different green points.
Therefore, we reran our Logit and hazard models with the levels of green
achievement as the focus variable, excluding all but the certified buildings
from the sample. Given the scores for Energy Star and points for LEED
were created on different scales, we rescaled and standardized them before
Our results are given in Table 7. In both the Logit and hazard models, green
points were significant and negative, meaning higher achievement earned
during certification was associated with lower default risk. The results suggest
the level of green achievement is salient to default risk. This finding also
alleviated the concern the green effect was due to uncontrolled locational
difference, as green building points are not likely to be correlated with any
locational advantage or disadvantage of a building.
Unfortunately, we could not tell from our analysis whether the effect from
higher green achievement on default risk was due to lower expenses, higher
rents or lower capitalization rates. All three are possible. Although Scofield
(2013) reported an inverse relationship between higher LEED ratings and site
energy use intensity, which suggests the inverse relationship between default
risk and green achievement could be explained by an inverse relationship
between green achievement and operating expenses, we do not know whether
operational improvements were made to the green buildings to qualify them
for green labeling. Therefore, we cannot say whether the decline in default
risk observed with increasing green achievement is best explained by a decline
in operating costs or an increase in demand in the space or capital markets.
Green Buildings in Commercial Mortgage-Backed Securities 23
Table 7 Impact of green points: logit and hazard model results.
Green Buildings Only
Logit Model Hazard Model
Green points 0.668*** 0.096**
Control variables Loan characteristics,
characteristics, nearby
transit density,
MSA-fixed effect and
vintage-fixed effect.
Contemporaneous LTV, current DSCR,
refinance incentive, macroeconomic
variables, loan characteristics,
building characteristics, nearby
public transit, MSA-fixed effect,
vintage-fixed effect and baseline
N709 21,263
Notes: (1) The focus variable “Green points” are the score or points received in the
certification process (standardized respectively). In the hazard model, they are time
varying as a building may get certified at a certain age of the loan after origination.
(2) *for p<0.1, ** for p<0.05 and *** for p<0.01.
A better data set with time series data on capital improvements, operating
expenditures and mortgage status is needed to assess these explanations.
Loan Terms
In addition to default risk, we wanted to know if terms for loans on green
building loans were different from those on nongreen buildings.
For about 86% of the certified buildings in our study, certification occurred
after the loan was originated (an average of five and six years after for Energy
Star and LEED, respectively). In these cases, we assumed the loan amounts
and terms at origination were the same as for otherwise similar buildings that
did not become certified (and we present evidence supporting this later).
Meanwhile, 14% of the green buildings in our study were certified (or in
the process) when their loans were originated. However, green premia may
not have been fully recognized in the underwriting process. Lending on com-
mercial buildings by national banks and federal savings associations follow
procedures specified by the U.S. Treasury’s Office of the Comptroller of the
Currency (Office of the Comptroller of the Currency 2013), which require
banks to adopt prudent underwriting standards. An independent state-certified
appraiser is required to establish the value for loans greater than $1 million
and by regulation (12 CFR 34.44) their appraisals must conform to the Uni-
form Standards of Professional Appraisal issued by the Appraisal Foundation.
24 An and Pivo
Those standards do not give any consideration to green buildings and it was
not until 2015 when the Appraisal Foundation issued its first voluntary guid-
ance for their valuation (Appraisal Foundation 2015). Moreover, an expert
committee on green buildings, convened in 2008 by the U.S. EPA, concluded
appraisers and lenders rarely possessed the data or knowledge to recognize the
value of green features, and often viewed them as insignificant or liabilities to
a building’s value (Choi 2009). Indeed, when the EPA panel was working, the
first studies on green premia were being completed. While we show below
this seems to be changing, we believe it was a fair characterization during the
time when many of the loans in our study were originated.
Therefore, we suspected the loan terms on green buildings were no different
than otherwise similar buildings. To test this, we compared buildings that
were and were not green registered or certified at loan origination on six
terms: mortgage interest spreads, IO periods, amortization periods, mortgage
constant and LTV. Our results showed some differences between the origi-
nal loan terms for green and nongreen buildings, but they were small and
economically negligible.
The CMBS loans we analyzed were originated during different time with
different maturity terms. Therefore, to meaningfully compare interest rates,
we focused on loans with the most popular maturity term, 10 years, and
calculated the mortgage spread for each loan. Spread was defined as the
mortgage interest rate minus the comparable maturity Treasury rate (risk-free
rate). Due to prepayments and defaults, the actual life of 10-year CMBS loans
is usually smaller than 10 years, therefore we used the 7-year Treasury bond
rate as the benchmark.20
Rate spread is potentially affected by various loan characteristics beyond
green certification. So, we conducted a regression analysis to test the impact of
certification on rate spreads in our sample of office buildings while controlling
for other variables.
The regression took the following form:
where riis the subprime mortgage spread for loan ioriginated at time t,Si
denotes a vector of green certification measures, Xi,tare control variables
including loan characteristics and market conditions and εiis the disturbance
20Alternatively, we used interest rate swap rate as the risk-free rate benchmark. The
results were robust.
Green Buildings in Commercial Mortgage-Backed Securities 25
Table 8 Impact of green label and certificate: rate spread regression.
Green at loan origination 0.115*** 0.146**
Green after loan origination 0.081*** 0.021
Green at loan origination
Green at loan origination
Green at loan origination
Control variables Loan characteristics, building characteristics,
nearby public transit, macroeconomic
variables, MSA-fixed effect and vintage-fixed
N5,354 466 466
Adjusted R-square 0.747 0.780 0.781
Note: (1) Complete model results for the full sample are in the Appendix; (2) *for p
<0.1, ** for p<0.05 and *** for p<0.01.
The control variables included the ones used for the Logit model: original
loan balance (in log terms), original LTV (a dummy for original LTV higher
than 70%), LTV and occupancy rate at loan origination, amortization term,
IO periods, whether there are prepayment restrictions such as lock out, pre-
payment penalty and yield maintenance, the age of the building (dummies
for less than or equal to 10 and 39 years), the per square foot property value,
yield slope curve, corporate credit spread, volatility of the 10-year Treasury
rate at loan origination and MSA- and vintage-dummies.
Remember, in our data we had certified buildings in two groups: those that
were certified at loan origination and those that were certified after loan
origination.21 Obviously, becoming green years after loan origination would
not affect the original interest rate. But we could use a variable for green
certification after loan origination to help control for unobservable differences
between green and nongreen buildings at the time of loan origination, as we
explain below.
We report our main regression results for mortgage spreads in Table 8 and
refer readers to the complete results in Appendix Table A2. In the full sample
21Among those certified at loan origination, 113 were Energy Star labeled, 7 were
LEED certified and 6 were LEED registered; so most of the certification effects
discussed here are attributable to Energy Star certification.
26 An and Pivo
(column 1 of Table 8), loans on office buildings green certified after loan
origination had rate spreads 8.1 bps lower than those that were never green.
But as mentioned earlier, earning certification after loan origination should
not have helped borrowers obtain better rates at loan origination, so we
think the difference reflects other unobservable differences between green
and nongreen buildings. However, we also see loans with green status at
loan origination had rate spreads 11.5 bps lower than those that were never
green, and the difference is bigger than between loans on buildings certified
after loan origination and those never certified (11.5 bps vs. 8.1 bps). We
believe the difference-in-difference (i.e., 3.4 bps) is the difference in the rate
spread caused by green certification at loan origination.
Using the matched-sample, we saw no significant difference between build-
ings that achieved green status after loan origination and those that never
did (column 2 of Table 8). Because we know our matching algorithm alle-
viated concern about unobservable variables, it is no wonder this result was
no longer significant. But the difference between buildings that were green
at loan origination and nongreen buildings remains significant (14.6 bps).
This echoes our prior finding with the nonmatched sample about the green
effect on mortgage interest rate at loan origination. However, we did notice
the size of the effect was economically insignificant (i.e., about $2,200 per
month or 1.4% in terms of monthly debt service based on the sample averages
for original loan balance and amortization term).22
Coefficients of the control variables shown in Appendix Table A2 largely fit
our expectations and findings in the existing literature. For example, LTV at
loan origination, amortization term, building age of 39 years or more, yield
slope, credit spread and Treasury rate volatility were all positively associated
with mortgage spread. Origination LTV higher than 70% was negatively
associated with mortgage spread, possibly due to a selection effect, where
lenders were willing to both accept a high LTV and low interest rate on very
good projects. This is consistent with findings in the existing literature (see,
e.g.,Anet al. 2013). The positive relation between lock out and prepayment
penalty clauses could be due to similar reasons, where lenders required those
clauses and higher interest rates on risky projects. There also were significant
vintage- and MSA-fixed effects.
Because the full and matched sample regression models did not account for
potential changes in demand for certification over time, we produced a third
22The average loan in our sample had an original loan balance of $24,252,000, interest
rate of 6.00%, and amortization term of 316 months, requiring $152,872 in monthly
debt service. Lowering the interest rate by 14.6 bps to 5.854% would lower the debt
service to $150,684. This is a savings of $2,188 per month or 1.4%.
Green Buildings in Commercial Mortgage-Backed Securities 27
Table 9 Impact of green label and certificate: IO periods, amortization term and
IO Periods
Ter m
Constant LTV
Green at loan origination 6.652*** 27.706*** 0.281*** 4.995***
Green after loan
4.897*** 0.012 0.172*** 5.369***
Control variables Loan characteristics, building characteristics, nearby
public transit, macroeconomic variables, MSA-fixed
effect and origination year- and month-fixed effect
N4,479 4,479 4,479 4,479
Adjusted R-square 0.670 0.509 0.579 0.248
Note: (1) For loans with 10-year term only; (2) *for p<0.1, ** for p<0.05 and
*** for p<0.01.
model using subperiods (column 3 of Table 8). The financial crisis subperiod
was missing because there were no loans originated then. The results are
quite interesting, suggesting lenders began lowering spreads on loans for
green buildings in 2006 and lowered them even more after the financial
crisis, perhaps as an indication of growing recognition of the green premia.
We also examined other loan terms that normally combine with the interest
rate to determine the cost of debt. Lenders can adjust these to manage their
risk and borrowers’ payments. We see that in our sample, for example, where
LTV at loan origination and the amortization term are related to rate spread
in the spread regression model (Appendix Table A2).
So, using regressions like Equation (2), we tested other dependent variables
including the IO term, amortization term and LTV ratio, again just using
loans with a 10-year maturity term. As shown in Table 9, we found office
buildings that were certified at loan origination had IO periods about 1.76
months (5.5%) longer than the mean for the full office sample (by compar-
ing the difference-in-difference between the two groups of green buildings
to the mean IO period given in Table 1). Using the same approach, we
found buildings that were green at origination had amortization terms about
27 months (8.5%) shorter than average and LTV ratios about a third of a
percent higher, which is virtually nominal in economic terms.
To give a simple illustration of how these terms work together to affect loan
payments, consider we said the reduction in rate spreads for buildings that
were green certified at loan origination could lower monthly payments on the
average original loan balance in the sample by about $2,200 (1.4%). However,
28 An and Pivo
if the amortization period on those buildings was on average 8.5% shorter,
the buildings that were green at certification would have paid $3,800 (2.5%)
more per month from the combined effect of the lower interest rate being
offset by the shorter amortization term.
Given the need to consider the terms together, we computed a mortgage
constant (per $100 loan balance) for each loan. To account for the impact of
the variable IO term, we calculated the mortgage constant as the weighted
average annual debt service per $100 loan balance using the IO period as a
percentage of the loan term as a weight. The mortgage constant was then used
as the dependent variable in a regression like Equation (2), to which we added
the loan origination year- and month-fixed effects to control for variations in
our base rate (7-year Treasury bond). The results are given in Table 9.
Using our difference-in-difference approach, we found the mortgage constant
was 0.11 lower for buildings green certified at loan origination. In comparison
to the mean mortgage constant in Table 1, that means the buildings certified
at loan origination paid 1.6% less per year in loan payments per dollar of
loan balance. These findings support the hypothesis that loan terms on green-
certified buildings are different than otherwise similar buildings, but they
show the effect is small enough to be economically negligible, especially
compared to the hazard rate reductions of 29% and 27% for Energy Star and
Energy Star or LEED, respectively, in our matched sample hazard model.
Conclusion and Discussions
We have shown green office buildings are associated with a decline in default
risk in CMBS loans. We controlled for potentially unobserved differences
between green and nongreen buildings by using longitudinal data in a hazard
model to observe the effect on risk before and after individual buildings
were certified under Energy Star or registered under LEED. We found being
Energy Star, or Energy Star and/or LEED, reduced default risk by 34%. We
then conducted a matched-sample analysis where the benchmark for each
green building was comparable buildings in the same ZIP code. Using the
matched-sample in the hazard models, we found similar green benefits: 29%
and 27% lower hazard rates for Energy Star and Energy Star or LEED,
respectively. These analyses support our general conclusion that loans on
green buildings outperform loans on nongreen buildings.
As for the possible mechanisms, we tested whether the lower default risk
could be explained by its effect on cash flow and equity rather than by the
labeling itself by using LEED points and Energy Star scores in the Logit and
hazard models. Both models support the hypothesis that higher points earned
Green Buildings in Commercial Mortgage-Backed Securities 29
in certification is associated with reduced default risk. We also tested possible
channels for the green effect and found it could be linked to an improved
equity position or contemporaneous LTV channel, in which lower default
risk reflects value appreciation in green buildings greater than for otherwise
similar properties.
Finally, we tested whether borrowers on green buildings enjoy better terms.
We did not expect that, based on evidence lenders and appraisers were not
considering green features when most of the loans in the study were origi-
nated. Using a regression to control other factors, we found being green at
loan origination was associated with better terms, though the effect was small
and economically negligible. We also found, however, lenders began lowering
spreads on loans for green buildings in 2006 and lowered them even further
after the financial crisis.
Of course, several questions remain open to further study. The exact mech-
anism through which green status affects mortgage default deserves closer
attention. Can the effect be fully explained by cash flow and equity channels,
or might other things be at play, such as differences between those who do
and do not own green buildings? There are also other important sustainability
features, aside from LEED and Energy Star status, that could be scrutinized
for their effect on mortgages such as walkability and transit-oriented devel-
opment. It would also be interesting to test these issues in other types of loan
pools outside the CMBS market.
Finally, our findings raise the question of whether underwriting tools and
practices should be amended to take advantage of these results and whether
lenders should offer more attractive terms on green buildings. Better terms
could improve overall market efficiency and environmental outcomes without
exposing lenders to greater risk.23
We thank two anonymous referees, Avis Devine, Nils Kok, Shimeng Liu,
Pat McAllister, Norm Miller, Paige Mueller, Joe Nichols, Doug Poutasse,
Andrew Sanderford, Jim Shilling, Chaoyue Tian, Bob White and participants
at the 2015 RERI Research Conference and the 2015 AREUEA National
Conference for their helpful comments. Financial support from the Real
Estate Research Institute (RERI) is gratefully acknowledged. We are also very
grateful to Real Capital Analytics, Redfin, Trepp and U.S. Green Building
Council for providing invaluable data. The views expressed in this article do
23By greater risk we mean compared to otherwise similar buildings. However, if
lenders give green buildings larger loans based on higher appraised values, then
benefits to default risk observed in this study would be reduced or disappear.
30 An and Pivo
not necessarily represent those of the Federal Reserve Bank of Philadelphia
or the Board of Governors of the Federal Reserve System. Any errors and
omissions are solely the authors’ responsibility.
Ambrose, B., M. Shafer and Y. Yildirim. 2016. The Impact of Tenant Diversification
on Spreads and Default Rates for Mortgages on Retail Properties. Journal of Real
Estate Finance and Economics
An, X., Y. Deng and S. Gabriel. 2016. Default Option Exercise over the Financial
Crisis and Beyond. SSRN Working Paper.
———, ———, J.B. Nichols and A.B. Sanders. 2013. Local Traits and Securitized
Commercial Mortgage Default. Journal of Real Estate Finance and Economics 47(4):
Appraisal Foundation. 2015. APB Valuation Advisory 6: Valuation of Green and
High-Performance Property: Background and Core Competency. Washington, DC:
The Appraisal Foundation.
Archer, W.R., P.J. Elmer, D.M. Harrison and D.C. Ling. 2002. Determinants of Mul-
tifamily Mortgage Default. Real Estate Economics 30(3): 445–473.
Baum, A. and A. McElhinney. 2000. The Causes and Effects of Depreciation in Office
Buildings: A Ten-Year Update. Working Papers in Land Management & Development.
07/00. Working Paper, University of Reading.
Carr, L.J., S.I. Dunsiger and B.H. Marcus. 2010. Walk Score as a Global Estimate of
Neighborhood Walkability. American Journal of Preventative Medicine 39(5): 460–
———, ——— and ———. 2011. Validation of Walk Score for Estimating Access
to Walkable Amenities. British Journal of Sports Medicine 45(14): 1144–1148.
Chen, J. and Y. Deng. 2013. Commercial Mortgage Workout Strategy and Conditional
Default Probability: Evidence from Special Serviced CMBS Loans. Journal of Real
Estate Finance and Economics 46: 609–632.
Cho, H., B.A. Ciochetti and J.D. Shilling. 2013. Are Commercial Mortgage Defaults
Affected by Tax Considerations? Journal of Real Estate Finance and Economics 46:
Choi, C. 2009. Removing Market Barriers to Green Development: Principles and
Action Projects to Promote Widespread Adoption of Green Development Practices.
Journal of Sustainable Real Estate 1(1): 107–138.
Ciochetti, B.A., Y. Deng, B. Gao and R. Yao. 2002. The Termination of Commer-
cial Mortgage Contracts through Prepayment and Default: A Proportional Hazard
Approach with Competing Risks, Real Estate Economics 30(4): 595–633.
———, ———, G. Lee, J.D. Shilling and R. Yao. 2003. A Proportional Hazards
Model of Commercial Mortgage Default with Originator Bias. The Journal of Real
Estate Finance and Economics 27(1): 5–23.
Das, P., A. Tidwell and A. Ziobrowski. 2011. Dynamics of Green Rentals over Market
Cycles: Evidence from Commercial Office Properties in San Francisco and Washington
DC. Journal of Sustainable Real Estate 3(1): 1–22.
——— and J.A. Wiley. 2014. Determinants of Premia for Energy-Efficient Design in
the Office Market. Journal of Property Research 31(1): 64–86.
Deng, Y., Z. Li and J. Quigley. 2012. Economic Returns to Energy-Efficient In-
vestments in the Housing Market: From Singapore. Regional Science and Urban
Economics, 42(3): 506–515.
Green Buildings in Commercial Mortgage-Backed Securities 31
Dermisi, S.V. 2009. Effect of LEED Ratings and Levels on Office Property Assessed
and Market Values. Journal of Sustainable Real Estate 1: 23–47.
Devine A. and N. Kok. 2015. Green Certification and Building Performance: Implica-
tions for Tangibles and Intangibles. Journal of Portfolio Management 41(6): 151–163.
Duncan, D.T., J. Aldstadt, J. Whalen, S.J. Melly and S.L. Gortmaker. 2011. Validation
of Walk Score for Estimating Neighborhood Walkability: An Analysis of Four U.S.
Metropolitan Areas. International Journal of Environmental Research and Public
Health 8: 4160–4179.
Eichholtz, P., R. Holtermans, N. Kok and E. Yonder. 2015. Environmental Perfor-
mance and Cost of Capital: Evidence from REIT bonds and Commercial Mortgages.
AREUEA 2015 Annual Conference Paper.
———, N. Kok and J. Quigley. 2010. Doing Well by Doing Good: Green Office
Buildings. American Economic Review 100(6): 2494–2511.
———, ——— and ———. 2013. The Economics of Green Building. Review of
Economic and Statistics 95(1): 50–63.
EPA. “Basic Information”. Green Building. U.S. Environmental Protection Agency,
20 Mar. 2016. Web. 6 Sept. 2017.
Episcopos, A., A Pericli and J. Hu. 1998. Commercial Mortgage Default: A Compar-
ison of Logit with Radial Basis Function Networks. Journal of Real Estate Finance
and Economics 17(2): 163–178.
Freybote, J., H. Sun and X. Yang. 2015. The Impact of LEED Neighborhood Certifi-
cation on Condo Prices. Real Estate Economics 43(3): 586–608.
Fuerst, F. and P. McAllister. 2009. An Investigation of the Effects of Eco-Labeling on
Office Occupancy Rates. Journal of Sustainable Real Estate 1: 49–64.
——— and ———. 2011. Green Noise or Green Value? Measuring the Effects of
Environmental Certification on Office Values. Real Estate Economics 39(1): 45–69.
Furfine, C.H. 2014. Complexity and Loan Performance: Evidence from the Securi-
tization of Commercial Mortgages. The Review of Corporate Finance Studies 2(2):
Goldberg, L. and C.A. Capone. 1998. Multifamily Mortgage Credit Risk: Lessons
from Recent History. Cityscape 4(1): 93–113.
——— and ———. 2002. A Dynamic Double-Trigger Model of Multifamily Mort-
gage Default. Real Estate Economics 30(1): 85–113.
Grovenstein, R.A., J.P. Harding, C.F. Sirmans, S. Thebpanya and G.K. Turnbull.
2005. Commercial Mortgage Underwriting: How well do Lenders Manage the Risks?
Journal of Housing Economics 14: 355–383.
Holterman, R. and N. Kok. 2017. On the Value of Environmental Certification in the
Commercial Real Estate Market. Real Estate Economics,
Jones, T. and G.S. Sirmans. 2016. Commercial Mortgage Termination and Pricing:
Theory and Evidence from 1970 to 2015. Journal of Real Estate Literature 2: 215–249.
Kaza, N., R. Quercia and C. Tian. 2014. Home Energy Efficiency and Mortgage Risks.
Cityscape 16(1): 279–198.
Kok, N., N. Miller and P. Morris. 2012. The Economics of Green Retrofits. Journal
of Sustainable Real Estate 4: 4–22.
Korniotis, G.M. and A. Kumar. 2013. State-Level Business Cycles and Local Return
Predictability. Journal of Finance 68(3): 1037–1096.
Mansfield, J.R. and J.A. Pinder. 2008. “Economic” and “Functional” Obsolescence:
Their Characteristics and Impacts on Valuation Practice. Property Management 26(3):
32 An and Pivo
McGrath, K.M. 2013. The Effects of Eco-Certification on Office Buildings: A Cap
Rates-Based Analysis. Journal of Property Research 30(4): 345–365.
Miller, N., D. Pogue and J. Saville. 2010. The Operations and Management of Green
Buildings in the United States. Journal of Sustainable Real Estate 2(1): 51–66.
———, J. Spivey and A. Florance. 2008. Does Green Pay Off? Journal of Real Estate
Portfolio Management 12(1): 1–12.
Newsham, G., S. Mancini and B.J. Birt. 2009. Do LEED-Certified Buildings Save
Energy? Yes, But . . . Energy and Buildings 41: 897–905.
Office of the Comptroller of the Currency. 2013. Commercial Real Estate Lending.
Washington, DC: Office of the Comptroller.
Pivo, G. 2013. The Effect of Sustainability Features on Mortgage Default Prediction
and Risk in Multifamily Rental Housing. Journal of Sustainable Real Estate 5(1):
———. 2014. Walk Score: The Significance of 8 and 80 for Mortgage Default Risk
in Multifamily Properties. Journal of Sustainable Real Estate 6(1): 187–210.
——— and J.D. Fisher. 2010. Income, Value, and Returns in Socially Responsible
Office Properties. Journal of Real Estate Research 32(3): 243–270.
——— and ———. 2011. The Walkability Premium in Commercial Real Estate
Investments. Real Estate Economics 39(2): 185–219.
Rauterkus, S.Y., G.I. Thrall and E. Hangen. 2010. Location Efficiency and Mortgage
Default. Journal of Sustainable Real Estate 2(1): 117–141.
Reichardt, A. 2014. Operating Expenses and the Rent Premium of Energy Star and
LEED Certified Buildings in the Central and Eastern U.S. Journal of Real Estate
Finance and Economics 49: 413–433.
———, F. Fuerst, N.B. Rottke and J. Zietz. 2012. Sustainable Building Certification
and the Rent Premium: A Panel Data Approach. Journal of Real Estate Research
34(1): 99–126.
Robinson, S. and P. McAllister. 2015. Heterogeneous Price Premiums in Sustainable
Real Estate? An Investigation of the Relation between Value and Price Premiums.
Journal of Sustainable Real Estate 7(1): 1–20.
———, R. Simmons, E. Lee and A. Kern. 2016. Demand for Green Buildings:
Office Tenants’ Stated Willingness-to-Pay for Green Features. Journal of Real Estate
Research 38(3): 423–452.
Scofield, J.H. 2013. Efficacy of LEED-Certification in Reducing Energy Consumption
and Greenhouse Gas Emission for Large New York City Office Buildings. Energy and
Buildings 67: 517–524.
Seagraves, P.A. and J.A. Wiley. 2016. The Cap Rate Spread: A New Metric for
Commercial Underwriting. Real Estate Economics 44(2): 490–520.
Seslen, T. and W. Wheaton. 2010. Contemporaneous Loan Stress and Termination
Risk in the CMBS Pool: How “Ruthless” is Default? Real Estate Economics 38(2):
Suzer, O. 2015. A Comparative Review of Environmental Concern Prioritization:
LEED vs Other Major Certification Systems. Journal of Environmental Management
154: 267–283.
Szumilo, N. and F. Fuerst. 2015. Who Captures the “Green Value” in the US Office
Market? Journal of Sustainable Finance & Investment, 5(1-2): 65–84.
Titman, S., S. Tompaidis and S. Tsyplakov. 2005. Determinants of Credit Spreads in
Commercial Mortgages, Real Estate Economics 33(4): 711–738.
——— and S. Tsyplakov. 2010. Originator Performance, CMBS Structures, and the
Risk of Commercial Mortgages. The Review of Financial Studies 23(9): 3558–3594.
Green Buildings in Commercial Mortgage-Backed Securities 33
U.S. Green Building Council. 2014. Getting to Know LEED: Building Operations and
Maintenance (O+M) [webpage]. Available at:
Vandell, K.D. 1992. Predicting Commercial Mortgage Foreclosure Experience. Jour-
nal of the American Real Estate and Urban Economics Association 20(1): 55–88.
Vandell, K., W. Barnes, D. Hartzell, D. Kraft and W. Wendt. 1993. Commercial
Mortgage Defaults: Proportional Hazards Estimation Using Individual Loan Histories.
Real Estate Economics 21(3): 451–480.
Vierra, S. 2016. “Green Building Standards and Certification Systems.” Whole Build-
ing Design Guide. National Institute of Building Science, 12-9-2016. Web. Accessed
Wiley, J.A., J.D. Benefield and K.H. Johnson. 2010. Green Design and the Market
for Commercial Office Space. Journal of Real Estate Finance and Economics 41:
Wu, P., Y Song, W. Shou, H. Chi, H.-Y. Chong and M. Sutrisna. 2017. A Compre-
hensive Analysis of the Credits Obtained by LEED 2009 Certified Green Buildings.
Renewable and Sustainable Energy Reviews 68: 370–379.
Yildirim, Y. 2008. Estimating Default Probabilities of CMBS Loans with Clustering
and Heavy Censoring. Journal of Real Estate Finance and Economics 37: 93–111.
Zhang, L., J. Wu, J. Liu and X. Zhang. 2017. The Value of Going Green in the Hotel
Industry: Evidence from Beijing. Real Estate Economics
Table A1 Complete results of the hazard model.
Estimate (S.E.)
Covariate Model 1 Model 2
Energy Star 0.411*
Energy Star or LEED 0.412**
Current DSCR 0.972*** 0.972***
(0.045) (0.045)
Contemporaneous LTV 0.466*** 0.466***
(0.044) (0.044)
Refinance incentive *lock out 0.279*** 0.28***
(0.084) (0.084)
Refinance incentive *no prepayment restriction 0.121 0.121
(0.086) (0.086)
Log loan balance 0.154*** 0.158***
(0.041) (0.041)
Origination LTV >70% 0.156*0.156*
(0.08) (0.08)
34 An and Pivo
Table A1 Continued.
Estimate (S.E.)
Covariate Model 1 Model 2
Public transit within 1/4 mile of the building 0.52*** 0.519***
(0.109) (0.109)
Value per sq. ft. 0.416*** 0.416***
(0.068) (0.068)
New building (age 10 years) 0.146*0.147*
(0.086) (0.086)
Old building (age >39 years) 0.104 0.104
(0.087) (0.087)
MSA unemployment rate innovation 0.392*** 0.392***
(0.056) (0.056)
Vol. of 10-year Treasury rate 0.226*** 0.226***
(0.051) (0.051)
Vol. of CPPI 0.029 0.029
(0.038) (0.038)
MSA-fixed effect Y Y
Vintage-fixed effect Y Y
N183,425 183,425
2LogL 15,358 15,358
AIC 15,448 15,448
Note:*for p<0.1, ** for p<0.05 and *** for p<0.01.
Table A2 Complete results of the spread regression.
Covariate Estimate S.E.
Energy Star labeled or LEED certified at loan origination 0.115*** 0.035
Energy Star labeled or LEED certified after loan origination 0.081*** 0.021
Log loan balance 0.032*** 0.007
Origination LTV >70% 0.045*** 0.011
Occupancy rate at loan origination 0.007 0.005
Amortization term 0.047*** 0.008
IO periods 0.033*** 0.009
Public transit within 1/4 mile of the building 0.013** 0.006
Value per sq. ft. 0.000 0.000
New building (age 10 years) 0.027*0.015
Green Buildings in Commercial Mortgage-Backed Securities 35
Table A2 Continued.
Covariate Estimate S.E.
Old building (age >39 years) 0.027** 0.013
With lockout term 0.100*** 0.035
With prepayment penalty 0.309*** 0.060
With yield maintenance term 0.003 0.016
Yield slope at loan origination 0.092*** 0.02
Credit spread at loan origination 0.065*** 0.011
Vol. of 10-year Treasury rate at loan origination 0.047*** 0.047
Loan term =5 years 0.529*** 0.05
Loan term =7 years 0.242*** 0.052
Loan term =10 years 0.080 0.047
Adjusted R-square 0.747
Note:*for p<0.1, ** for p<0.05 and *** for p<0.01.
Table A3 Tests of different Energy Star label definitions.
Temporary Effect
Permanent Effect
Energy Star labeled 0.411** 0.427** 0.336**
(0.663) (0.652) (0.715)
Control variables Contemporaneous LTV, current DSCR, refinance
incentive, macroeconomic variables, loan
characteristics, building characteristics, nearby public
transit, MSA-fixed effect, vintage-fixed effect and
baseline hazard
Notes: (1) Hazard ratio in parenthesis; (2) the exact same specification as in Appendix
Table A1 except the definition of Energy Star label definitions are different; (3) in the
current definition, if a building is Energy Star labeled, we assume the effect will last
at least in the next year even if the label is not renewed in the second year; in the
temporary effect definition, we assume the effect is only in the year when the label is
obtained (renewed); and in the permanent effect definition, we assume once a label is
obtained its effect will last for the whole life of the loan; (4) *for p<0.1, ** for p<
0.05 and *** for p<0.01.
36 An and Pivo
Table A4 Test of higher-order effects of current DSCR and LTV.
Covariate Estimate S.E.
Energy Star 0.420** 0.192
Current DSCR 0.930*** 0.079
Current DSCR squared 0.013 0.037
Contemporaneous LTV 0.809*** 0.077
Contemporaneous LTV squared 0.077*** 0.014
Control variables Y
MSA-fixed effect Y
Vintage-fixed effect Y
2LogL 15,326
AIC 15,420
Note:*for p<0.1, ** for p<0.05 and *** for p<0.01.
... This will promote the development of green building via the application of GF instruments such as green fiscal investment, green credit, green insurance, and green bonds (He et al., 2022). Increased green building related research that focuses on the debt market (An & Pivo, 2020) will help developers and investors to understand the present GF challenges and opportunities in green building. Again, in using real case examples, favourable economic (investment returns) and lower interest rates associated with GF (Akomea-Frimpong et al., 2022) can be modelled to assess the economic feasibility of GF in green building projects (Taghizadeh-Hesary et al., 2022). ...
... Additionally, to increase GF in green building econometric studies and archival data analysis, improved databases on GF in green building is necessary. Such databases could promote research on the impact of green building finance on carbon emissions (Gholipour et al., 2022), economic feasibility studies (An & Pivo, 2020), and time series econometrics (He et al., 2022) of GF in green building at both national and international levels. Besides, the challenges associated with analyzing investment risks of GF in green building could be reduced with adequate data (Akomea-Frimpong et al., 2022). ...
Green finance (GF) supports the global fight against climate change and its impacts. It is critical to attaining the Paris Agreement and the United Nations Sustainable Development Goals. Since GF is regarded as the future of finance and investment, it needs to be fully understood. This paper presents the first mixed-methods systematic review with both bibliometric and qualitative analysis of the state-of-the-art and trends in GF research. A bibliometric review was performed to quantitatively examine the main areas of interest, journals, and clusters of GF research based on 995 related publications retrieved from Scopus and validated with the Web of Science, Google Scholar, and ScienceDirect. Results showed that GF is still relatively an immature but interdisciplinary research area. A further qualitative-systematic analysis of 60 selected publications was conducted to identify the key findings, challenges, and recommendations for future research. Findings revealed six major research hotspots in GF: (i) green bond market and greenium, (ii) green credit (loan), (iii) carbon investment and market, (iv) green banking, (v) market stress (e.g., the COVID-19 pandemic) and GF, and (vi) domestic and international climate finance policies. Based upon gaps in extant literature, suggestions for future research are proposed: GF policy initiatives and incentives; GF in green building; and Fintech-for-GF. This study provides insights into key applications of GF as it applies to specific research fields, as well as the pathways to realize the accruable benefits of GF to enhance research and development.
... Some believe that environmental performance is related to the effective use of resources for the overall improvement of organizational efficiency [21], while others believe that environmental performance will improve corporate image and reputation, which may help enterprises gain benefits in the labor, commodity and capital markets [22]. Other scholars believe that investment in environmental performance will reduce the operational risk of enterprises, and the reduction of operational risk can make it easier for enterprises to obtain funds and even reduce financing costs [23,24]. Thompson et al. through questionnaires and semi-structured interviews with the Bank of England, which granted corporate loans, investigated that the public's attention to environmental information has become the motivation for banks to pay attention to corporate environmental information, and banks will use the environmental performance information in the corporate annual report when making loan decisions [25]. ...
Full-text available
High environmental performance of enterprises may reduce financing costs, while good environmental performance can promote sustainable development of enterprises. Therefore, this paper examines the impact mechanism of China’s corporate environmental performance on financing costs, and whether corporate sustainable development plays a regulatory role in the research of heavy pollution industries. This study is conducted through the Breusch and Pagan Lagrange multiplier test for random effects and the Hausman test to determine whether to adopt Fixed-effects regression or Random-effects GLS regression as an estimation method to control individual effects and endogenous problems brought by time. By collecting the samples of listed companies in China from 2010 to 2021, the empirical results show that corporate environmental performance is negatively related to financing costs. Sustainable development, as a moderator variable, is negatively related to financing costs and has weakened the inhibition of corporate environmental performance on financing costs. Although the existing literature shows that environmental performance will lead to changes in debt costs, this study has made contributions to the literature by revealing the sustainable development mechanism in the relationship between corporate environmental performance and financing costs and has verified that sustainable development is one of the important factors affecting financing costs.
... However, little empirical research has been conducted on the relationship between energy efficiency and loan pricing. An and Pivo (2018) examined the default risk and loan terms of loans for energycertified buildings in the US office building market. They found that with regard to loans granted for buildings that were already green-certified upon loan origination, banks set interest rates that were 15 basis points lower on average than for buildings that were green-certified only after the loan was granted. ...
The real estate market is a key component of the green transition, and thus it is worth examining the pricing and financing costs of modern residential buildings in Hungary. In our study, we investigate a) whether a significant price premium can be identified for green properties based on new housing projects in Budapest, and b) whether banks finance residential buildings with more advanced energy technologies at lower interest rates. Based on our regression estimation, the green price premium is clearly evident in the Budapest new housing market: on average, homes with an energy rating of BB or better are 5.1 per cent more expensive than homes with an energy rating of CC. Based on our estimate of housing loan interest rates, no significant difference can be identified in the interest rates on loans granted to finance properties with a certificate higher than CC compared to loans granted for properties with CC certificate, i.e. banks do not yet factor energy aspects into the pricing of loans.
... Such risks affect the calculation of probabilities of default on loans and debt service coverage ratios. The evidence reported by An and Pivo (2017) for the LEED and ENERGY STAR certification programmes in the United States indeed shows that green buildings carry lower default risk, all else being equal. ...
The real estate sector, including the residential and commercial market segments, is a heavy consumer of energy and, as a result, a sizeable source of emissions of greenhouse gases. This is primarily on account of the consumption of energy in heating and cooling systems, as well as in the use of domestic appliances. The construction, maintenance and thermal characteristics of buildings add to the sector's energy consumption. Based on a review of scholarly and policy-focused work, this paper argues that decarbonisation strategies to meet agreed climate change targets will need to incorporate policies targeted to the specificities of the real estate sector. They include addressing split incentives among owner-occupiers, landlords and renters (in the private and social housing markets) for investment in home improvements and energy retrofitting; raising the standards of energy performance for new and existing properties through labelling/certification and other means; and reducing the cost of finance for needed investments while broadening access to the underserved population.
Az ingatlanpiac kardinális eleme a zöld átállásnak, ezért érdemes megvizsgálni, hogy milyen árazással és finanszírozási költséggel érhetők el a korszerű lakóingatlanok Magyarországon. Tanulmányunkban egyrészt azt vizsgáljuk, hogy kimutatható-e szignifikáns árprémium a zöld ingatlanokra a budapesti új építésű lakásprojektek alapján, másrészt hogy a bankok alacsonyabb kamatszint mellett finanszírozzák-e a korszerűbb energetikával felszerelt lakóingatlanokat. Regressziós becslésünk alapján a zöld árprémium egyértelműen kimutatható a fővárosi új lakások piacán: a BB vagy jobb energetikai besorolással rendelkező lakások átlagosan 5,1 százalékkal drágábbak, mint a CC energetikai besorolású lakások. A lakáshitelek kamatlábára vonatkozó becslésünk alapján a CC-nél magasabb tanúsítvány kategóriájú ingatlanok finanszírozására nyújtott hitelek kamatlábára vonatkozóan nem azonosítható szignifikáns eltérés a CC tanúsítvánnyal rendelkező ingatlanokra nyújtott hitelek kamatlábához képest, azaz a bankok jelenleg még nem érvényesítik az energetikai szempontokat a hitelek árazásában.
This article presents a systematic literature review of the determinants of commercial real estate loan credit spreads from 1989 to 2019. The motivation for this analysis is twofold. First, credit spread models receive little attention in commercial real estate, since research mainly focuses on the drivers of housing credit spreads. Second, data availability in the commercial real estate loan sector is still scarce and the data often mixed with other asset or loan types, such as commercial mortgage-backed securities and their related determinants. We synthesize the literature into four dimensions: property and loan characteristics, macroeconomic aspects, and counterparty characteristics. Using an open coding process, the literature is thoroughly analyzed in 10 subcategories. The results disclose that the main determinants are property and loan characteristics, which impact the reliability and riskiness of cash flow and interest and debt service. Since researchers focus primarily on credit risk determinants, counterparty characteristics have lately gained more attention as credit spread determinants, which deserves additional empirical research. Furthermore, we observe that endogeneity issues have become more prevalent, despite difficulties in finding the best treatment and different impacts.
This chapter organizes the review of our perspectives on Green Market Transformation into factors that predict and explain Green Market Transformation. First, we describe the conditions associated with this transformation. These are factors that supports the transformation occurring. To transform markets through voluntary mechanisms requires the strong rule of law, institutions to curtail fraud and to support property rights to encourage long-term investments. These boundary conditions become assumptions about the context that can help enable Green Market Transformation. Next, the catalysts are reviewed – those causal mechanisms that actually bring about the deep greening of the building sector markets – that previous chapters detail. Our theory of Green Market Transformation argues that the dissemination of information through voluntary mechanisms can help address the barriers and market failures that produce the Valley of Death. It is hypothesized that these mechanisms include, though are not limited to, well-designed ecolabels and iterative demonstration projects that overcome market barriers by disseminating information about the cost and performance of nascent technologies to the marketplace, lowering costs and risks associated with the adoption of greener technologies. These predictive conditions and causal mechanisms form the blueprint for those seeking to engineer green market transformation.
This chapter reviews potential concerns of green building, including the environmental impact of the buildings, equity impacts, and environmental justice implications of ecolabeled buildings. These concerns typically revolve around the ultimate environmental impacts of green buildings and the equity implications of how we transition to a greener built environment. Green buildings may not be as green as we expect or want, and price premiums for green buildings work against affordability. Our green market transformation story is not a naïve, romantic, idealized story of perfectly sustainable practices overtaking our foolish old ways. This transformation story is messy, fraught with imperfections, and leaves ample room for improvement. In fact, that is part of the essence of this story: Iterative, ongoing improvements, building momentum toward a more sustainable system. Openly drawing our attention to these concerns and shortcomings can help us turn them into opportunities for continued gain and building on that momentum. Market transformation does not happen overnight, and it does not stop after a singular change. It is an ongoing evolution. This chapter reviews some of the shortcomings and concerns about this otherwise positive evolutionary path for green buildings.
Ecolabeled green buildings can have a diverse array of characteristics. Their superior environmental performance can include things like energy efficiency or water efficiency; using cleaner and lower carbon energy sources; sourcing construction materials with sustainable practices; or site selection for the buildings so as to reuse and rehabilitate brownfields or encourage use of public transportation or bicycles. The multidimensional nature of “greenness” for green building is an essential feature of these ecolabels and the Green Building Movement more broadly. The holistic approach to greener buildings embraces flexibility, diversity, and innovation over strictly prescriptive or one-size-fits-all approaches. In this chapter we unpack the diversity of green buildings using attributes such as the publicness and the private marketing benefits of an organization’s ecolabeling strategy to provide improved understanding of the manner in which firms certify green. Green building strategies are classified as altruist, pragmatist, green club, and greenwash. By providing better understanding of green building strategies, our understanding of sustainability strategies by firms and organizations is enhanced.
Options trading on Real Estate Investment Trusts (REITs) has grown exponentially over the past two decades. This article investigates whether, and to what extent, higher levels of active options trading activity materially influence REIT financial market performance. Using a sample of 224 equity REITs, over the period 1996 through 2019, we found REITs with higher levels of options trading activity persistently exhibit enhanced valuations and financial market liquidity. We further found the positive impact of options trading activity on REIT value is not driven by the positive association between options trading activity and market liquidity, but rather is primarily attributable to the enhanced informational efficiency induced by options trading. Taken together, these results provide compelling evidence of the potential benefits of options trading for publicly listed REITs.
Full-text available
This paper investigates the effect of eco-labeling on the occupancy rates of commercial offices in the United States. The occupancy rates of LEED and ENERGY STAR-labeled offices are compared to a sample of non-labeled offices. Using OLS and quantile regression analyses, a significant positive relationship is found between occupancy rate and the eco-label. Controlling for differences in age, height, building class, and quality, the results suggest that occupancy rates are approximately 8% higher in LEED-labeled offices and 3% higher in ENERGY STAR-labeled offices. However, for ENERGY STAR-labeled offices, effects are concentrated in certain market segments.
Full-text available
Based on several unique datasets in Beijing, this paper investigates the value of going green in the hotel industry by combining the traditional hedonic pricing model with the state-of-the-art content analysis of online reviews. The results indicate that, the rate of complaints about the indoor environmental quality of green hotels is roughly 19% lower than that for non-green hotels. Hedonic regression analysis concludes that green hotels enjoy a significant room rate premium of 6.5% without reducing occupancy rates, mainly due to improved indoor environmental quality. Recognizing the presence of such co-benefits is likely to induce hoteliers to embrace green practices. This article is protected by copyright. All rights reserved
Full-text available
A significant part of the global carbon externality stems from the real estate sector. Environmental certification is often hailed as an effective means to resolve the information asymmetry that may prevent markets from effectively pricing the energy performance of buildings. This study analyzes the adoption and financial outcomes of environmentally certified commercial real estate over time. We document that nearly 40 percent of space in the 30 largest U.S. commercial real estate markets holds some kind of environmental certification in 2014, as compared to less than 5 percent in 2005. Tracking the rental growth of 26,212 office buildings, we measure the performance of environmentally certified real estate over time. We document that certified office buildings, on average, have slightly higher rental, occupancy and pricing levels, but do not outperform non-certified buildings in rental growth over the 2004-2013 period. Further performance attribution analysis indicates that local climate conditions, local energy prices and the extent of certification lead to significant heterogeneity in market pricing. On aggregate, these findings provide some evidence on the efficiency of the market in the adoption and capitalization of environmental characteristics in the commercial real estate market.
In this paper, I use logistic regression to study the relationship between walkability and mortgage default risk in multifamily housing in a pool of nearly 37,000 Fannie Mae loans. Walkability is measured with Walk Score, a widely available metric. Controls were introduced for loan terms, property characteristics, neighborhood conditions, and macroeconomics. Walkability reduced default risk but the relationship was nonlinear with thresholds. Default risk significantly increased where walkability was very low and significantly decreased where it was very high. The implication is that walkability and its possible benefits to health and the environment could be fostered by relaxing lending terms without adding default risk.
This is the first study focused on the economics of green renovations. Our findings are focused on Leadership in Energy and Environmental Design (LEED) buildings certified under the Existing Building: Operations and Maintenance (EBOM) certification scheme during the 2005-2010 period. We compare rents and occupancy rates, and investigate the types of improvements undertaken, as well as the amount of investments required. We survey building owners on the typical improvements and their attitudes towards the benefits and costs of upgrades. The findings indicate that investments in ‘‘green’' retrofits are incorporated by the market, which is consistent with past studies that mostly focused on new construction. The findings indicate that, on average, investments in the sustainability of commercial buildings are economically viable.
This study examines the relationship between transportation-, location-, and affordability-related sustainability features and default risk in multifamily housing. It finds that sustainability features can be used to improve the prediction of mortgage default and reduce default risk. The study uses 37,385 loans in the Fannie Mae multifamily portfolio at the end of 2011:Q3. The results suggest two implications for practice. First, certain aspects of sustainability can be fostered without increasing default risk by adjusting conventional lending standards. Second, lenders could improve their risk management practices by taking stock of sustainability features when loans originate.
This study examines the performance of green buildings from the operation and management perspective. Specifically, we look at the utility expenses, cleaning practices, use of energy-saving devices, and other building operation procedures of a national sample of office properties managed by CB Richard Ellis. The findings indicate that green buildings in the sample are more energy-efficient than their non-green counterparts. Surprisingly, the average total operating expenses of the green building group is higher than the non-green building group, albeit insignificantly. Additionally, a building's operating performance is more highly correlated with its ENERGY STAR score, and not the ENERGY STAR label.
The increasing societal focus on environmental issues leads to important questions about the relationship between corporate environmental (ESG) performance and firms’ cost of capital, but research on this topic remains scant. The real estate sector offers an ideal testing ground to investigate the relationship in two distinct manners, while specifically addressing concerns about endogeneity. We first investigate the spreads on commercial mortgages collateralized by real assets, some of which are environmentally certified. We then study spreads on corporate debt of property companies (REITs), both at issuance and while trading in the secondary market. The results show that loans on environmentally certified buildings command lower spreads than conventional, but otherwise comparable buildings, varying between 24 and 29 basis points, depending on the specification. At the corporate level, REITs with a higher fraction of environmentally certified buildings have lower bond spreads in the secondary market. These results are robust to different estimation strategies, and signal that environmental risk is efficiently priced in the real estate debt market.
This research examines the effects of energy efficiency certification levels on office rental rates and lease structures to determine whether any cost benefits of green buildings are captured by landlords or remain, at least partially, with the tenant. To this aim, our analysis applies the largest and most detailed data set to date, a panel of 14,283 US office properties. Using fixed-effects and dynamic Arellano–Bond frameworks allows us to estimate the differential rental price impact of Energy Star certification both across and within buildings. The general results indicate that buildings with higher levels of energy efficiency achieve higher gross rents allowing landlords to benefit from the premium. However, improved energy efficiency over time is also linked to a slower growth of rental prices as some of the benefit is passed onto tenants. Interestingly, the cost-saving benefit of energy efficiency appears to have the strongest impact on rental rates.