Energy Factors, Leasing Structure and the Market Price
of Office Buildings in the U.S.
∗
Dwight Jaffee
†
, Richard Stanton
‡
and Nancy Wallace
§
August 31, 2011
Abstract
This paper presents an empirical analysis of the relationship between energy factor
markets, leasing structures and the transaction prices of office buildings in the U.S.
We employ a large sample of 15,133 office building transactions that occurred between
2001 and 2010. In addition to building characteristics, we also include information on
the operating expenses, the net operating income, and the market capitalization rates
at sale to estimate an asset pricing model for commercial office real estate assets. A
further set of important controls in our analysis is the one-to-twelve month forward
contract prices and the shape of the forward contract price curve, using auction data
for the regional electricity trading hubs in which the building is located and auction
data from the Henry Hub for natural gas. We also include weather metrics in the form
of the variance in the last twelve months of minimum and maximum temperature and
precipitation for each building’s location and sale date. Our final set of controls in-
cludes information on the dominant contractual leasing structure of the buildings. Our
empirical results suggest that Energy Star labels do not explain additional variance
in property prices once the key asset pricing factors of expenses, income and market
capitalization rates are included. Energy factor market prices, the shape of the energy
forward price curves, and weather metrics are consistently shown to be statistically sig-
nificant determinants of office building transaction prices, suggesting that commercial
office building prices are likely to be exposed to shocks in these markets.
∗
This work was supported by the Assistant Secretary for Energy Efficiency and Renewable Energy,
Building Technologies Program, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231.
†
‡
§
Disclaimer
This document was prepared as an account of work sponsored by the United States
Government. While this document is believed to contain correct information, neither the
United States Government nor any agency thereof, nor The Regents of the University of
California, nor any of their employees, makes any warranty, express or implied, or assumes
any legal responsibility for the accuracy, completeness, or usefulness of any information, ap-
paratus, product, or process disclosed, or represents that its use would not infringe privately
owned rights. Reference herein to any specific commercial product, process, or service by
its trade name, trademark, manufacturer, or otherwise, does not necessarily constitute or
imply its endorsement, recommendation, or favoring by the United States Government or
any agency thereof, or The Regents of the University of California. The views and opinions
of authors expressed herein do not necessarily state or reflect those of the United States
Government or any agency thereof or The Regents of the University of California.
Contents
1 Introduction 1
2 Commercial Office Building Market Value 4
3 The CoStar Data 6
3.1 Energy Star Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Local Weather Data Matching . . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.3 Energy Auction Data Matching . . . . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Summary Statistics for the CoStar Transaction Data . . . . . . . . . . . . . 11
4 The Investor’s Energy Star Rating Decision 16
5 Empirical Results with Transactions Prices 20
6 Leases 25
6.1 Lease Contracts and Energy Efficient: Economic Theory . . . . . . . . . . . 28
6.1.1 Contract Theory, Uncertainty, and Energy Use . . . . . . . . . . . . . 28
6.1.2 Capital Investments Under Alternative Lease Contracts . . . . . . . . 30
6.2 Empirical Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
7 Conclusions 35
A Appendix 36
A.1 Geographic Structure of the Transactions Data . . . . . . . . . . . . . . . . . 36
A.2 Weather Data Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
A.3 Energy Data Construction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
A.3.1 Forward Market for Power (Electricity) . . . . . . . . . . . . . . . . . 37
A.3.2 Dataset - Platts-Ice Forward Curve . . . . . . . . . . . . . . . . . . . 39
A.3.3 Futures Market for Natural Gas . . . . . . . . . . . . . . . . . . . . . 41
1 Introduction
In the United States, buildings consume almost 39% of the total energy consumption, making
real estate the largest energy consuming sector by a considerable margin. By comparison as
shown in Table 1, U.S. industrial consumption of energy is about 33% of the total and the
transportation share is about 28% of the total. Within the real estate sector, the share of
total energy used is almost equally split between residential (20.9%) and commercial (18.0%)
buildings.
Table 1: Buildings Share of U.S. Primary Energy Consumption (Percent), 2006
Residential
Buildings
Commercial
Buildings
Total
Buildings
Industry Transportation Total Total
Consumption
(quads)
20.9% 18.0% 38.9% 32.7% 28.4% 100% 99.5
Source: U.S. Department of Energy (2009)
An additional sobering fact is that the energy consumption of U.S. real estate appears, to
be substantially less efficient than comparable European buildings, even after controlling for
such factors such as climate, GDP, and population.
1
There is no mystery as to the reason:
U.S. costs for electricity and heating oil range from 50 to 75 percent of the levels in most
European countries (see ?). This indicates there must be feasible technologies that would
allow the energy consumption of U.S. buildings to be reduced significantly. Such investments
could arise as the result of building codes and comparable requirements, or as a voluntary
response to high and volatile energy prices.
Most economists would agree that the “first-best” solution to reducing U.S. energy con-
sumption, across the three major uses of real estate, industry, and transportation, is to raise
the U.S. price of energy through appropriate fiscal instruments. While this may occur in
the future, the timing of this change in policy remains highly uncertain. So it is sensible,
even critical, to look for alternative, accelerated, mechanisms to reduce U.S. energy con-
sumption. Indeed, if and when U.S. energy prices do rise, the economic adjustment will be
easier and faster if the transformation toward a more energy efficient technology is already
well along. Facing inadequate energy-efficient investments in real estate, governments in
both the U.S. and Europe have intervened to provide additional incentives. To date, three
1
Two reports by the International Energy Agency (see ??) provide comparisons of residential energy use
in the U.S. and Europe corrected for climate and measured per unit of GDP or per capital. ? also shows
substantially higher U.S. energy consumption compared to Europe after controlling for GDP and population.
? provides a discussion comparing energy use in the U.S., Australia and the European Union.
1
approaches have dominated theses interventions: expanded building codes, energy-efficiency
certifications, and direct fiscal subsidies.
Expanded building codes have been the primary mechanism to ensure energy-efficient
structures in Europe and are likely to expand over time in the U.S. as well. Building code
requirements, however, have their primary impact on new construction, and given the long
durability of most buildings, new construction annually represents a very small percentage
of the existing building inventory. Most building codes, furthermore, are prescriptive rather
than performance-based, which limits the incentive they provide for new innovative solutions.
Disclosure certificates at the time of building sales have also become an important mecha-
nism in Europe. The 2002 EU Energy Performance in Buildings Directive requires an energy
performance certificate based on either the building’s design or usage characteristics. While
such disclosures might well have an impact on the sales prices, it is unclear to what extent
such disclosures motivate either the seller or buyer to initiate energy efficient investments. In
the U.S., certifications such as Leeds and Energy Star are available, and such certifications
are sought by builders and developers who are planning to create energy efficient buildings.
But these certifications primarily apply to new buildings, and it is unclear to what extent
the certifications are the factor actually motivating energy-efficiency in these buildings.
2
Direct subsidies provided by either government agencies and public utilities may provide
economic stimulus to energy efficient investments. In these times of harsh fiscal budgets,
however, direct subsidies are unlikely to be a major driver of new energy-efficient investments.
Given the limitations of building code requirements, energy-efficiency certifications, and
direct subsidies, it is critical to consider whether there are other solutions to the private
market failures that are contributing to the under-investment in U.S. building energy effi-
ciency. It is possible, of course, that the available technology is just not profitable in net
present value (NPV) terms. However, it appears that NPV positive energy-efficient invest-
ment opportunities already exist, and that further innovations and cost-savings from scale
economies are literally in process.
In this paper, we focus on the relationship between energy efficiency and commercial office
building transaction values. Our analysis is based on a comprehensive data set of commercial
office building transactions. The data set was developed to include a rich set of controls for
heterogenous building characteristics such as each building’s capacity, quality, utilization,
and lease contract structure and matched building-specific market information such as the
2
? show that buildings with certifications obtain higher sales prices, which provides a motivation for
obtaining such certifications. However, it is unclear whether the existence of the certifications motivates
energy efficient investments that otherwise would not have been carried out. The existence of higher sales
prices on certified buildings also does not indicate whether the initial energy-efficiency investments were
NPV positive; to show that, the investments costs must be compared with the sales prices.
2
actual capitalization rate at sale, local-level wholesale energy market price dynamics and
local weather pattern dynamics at the time of sale. Our goal is determine whether there
exists a statistically significant empirical relationship between the energy performance of
commercial office buildings and their transaction values controlling for market and building
characteristics.
Our sample includes office buildings found in the CoStar data that were located in
U.S. office markets with 150,000 or more employees working in the employment category
Information, Finance, and Professional and Business Services (the major office categories).
3
We developed time and market-specific local weather station data and energy forward con-
tract auction data from the trading hubs appropriate to each building and linked these data
to each useable transaction record in the CoStar data. Our transaction data set is comprised
of office building sales between 2001 and 2010. The data set includes 15,133 arms-length
4
transactions for office buildings located in 43 U.S. metropolitan areas. Importantly these
data are suitable to analyze the relationship between the market transaction values of U.S.
office buildings and their structural, contractual, energy and market-related characteristics.
A second objective of the paper is to determine whether U.S. Environmental Protection
Agency (EPA), Energy Star Certification, certification that a building operates in the top
quartile of energy efficiency in the U.S., affects transaction values. To meet this objective,
we obtained information on the population of Energy Star rated buildings from the EPA.
The EPA list included the exact location of the building, the numeric Energy Star rating,
and the date at which the rating was granted. These data were then merged with the CoStar
transaction data so that we can identify all buildings that had an Energy Star rating prior
to their sale date. Finally, for a subset of the transactions we have complete information on
each building’s total operating expense, gross and net operating income, and capitalization
rate at the sale date along with information on the first and second mortgage debt structure.
This paper is the first product in a research program that is designed to develop foun-
dational relationships between real estate asset prices, real estate energy consumption, and
the energy efficiency price risk exposure of commercial real estate in the U.S. Our goal in
this paper is to establish the link between commercial real estate asset prices, operating
expenses, and energy factor market inputs. In a companion paper, we seek to establish the
link between the energy risk exposure of a building and the valuation of commercial real
estate mortgages that also share this risk exposure. The two parts of this research program
are needed to develop suitable valuation technologies to underwrite the energy efficiency risk
3
Bureau of Labor Statistics, Employment Hours and Earnings, State and Metro Area, http://www.bls.
gov/sae/data.htm.
4
These are sales between unrelated persons
3
of commercial real estate mortgages. Energy risk pricing technology for commercial real
estate mortgage valuation is also needed to allow for the development of new bond markets
to support the capital requirements for energy efficiency retrofits in the U.S. Finally, this
work should provide better risk metrics for the existing commercial real estate mortgage
market that currently does not underwrite/or price the energy components of commercial
real estate mortgage default risk.
The paper is organized into seven sections. In the next section, Section 2, we discuss
empirical representations for the value of commercial real estate. In Section 3, we provide
details on the construction of our data sets. In Section 4, we conduct an empirical investi-
gation of the Energy Star rating decision of investors in commercial real estate in the U.S.
Section 5 presents empirical results for tests on the relationship between real estate trans-
action prices, operating expenses, net operating income, the capitalization rate, and Energy
Star labels. Section 6 considers the role of leasing structure on the pricing of commercial
real estate assets and the contractual management of energy price passthroughs to tenants.
Section 7 concludes.
2 Commercial Office Building Market Value
The canonical representation for the market value of a commercial real estate asset is as the
discounted present value of the asset’s future net operating income. Since well maintained
office properties typically can be assumed to be long-lived assets, the market price of a
commercial office building at the investor’s purchase date (t = 0) can be written as
P
0
=
∞
X
t=1
R
t
(1 + i
t
)
t
(1)
where P
0
is the market price at the investment date, t = 0, R
t
is the net operating income at
the t
th
period. The forward path of the net operating income is defined as the forward path
of gross effective income (rent per square foot times the square footage rented minus the
vacancy rate) minus the forward path of the total operating expenses. The market interest
rate, i
t
, at the t
th
period is defined as the riskless rate plus a risk premium. Since there is
rarely sufficient information about the future path of net operating income for these assets,
often the current net operating income is used as a sufficient statistic for the future net income
given an assumption about the market growth rate for this income stream. Assuming a flat
4
term structure, the valuation formula for a commercial real estate asset can be written as,
P
0
=
∞
X
t=1
R(1 + g)
t
(1 + i
t
)
t
=
R
(i − g)
(2)
where g is the market growth rate for net operating income and (i − g) is known in the real
estate industry as the market capitalization rate, where it is assumed that i − g > 0.
In a recent paper, Plazzi, Torous, and Valkanov (2010) find that metropolitan-level
macroeconomic conditions appear to significantly impact the dynamics of the observed
growth rates in net operating income across regional markets. In addition to these growth
rate dynamics, macro-economic shocks to factor input prices, such as energy and labor costs,
are also likely to affect the level of net operating income, and thus the level of asset prices,
through total operating expenses. An important outstanding question is the degree to which
shocks to the energy factor can be mitigated by building characteristics such as the energy
efficiency of building engineering systems, such as heating, air conditioning and ventilation,
or low energy-use lighting systems. To our knowledge, other than engineering simulation
studies, there are no systematic empirical analyses of the relative operating efficiency of of-
fice buildings in U.S. metropolitan markets. The primary impediment to such studies is the
lack of information on office building transaction prices along with information on building-
level total operating expenses, gross/net operating expenses, capitalization rates, location,
energy-use metrics, and physical characteristics of the building.
In an influential recent paper ? measure the economic value of the certification of “green
buildings” that have either received an Energy Star rating or a LEED rating. They find that
these buildings have asking rents (prices) that are about 3% (16%) higher per square foot
than the asking rents (prices) found in otherwise spatially identical buildings. Two other
recent studies by ? and ? also find higher asking rents, prices, and occupancy rates for
buildings with green ratings. All three of these studies test for the effects of “green” ratings,
using a hedonic representation of prices (asking rents) as a function of the characteristics and
location of the building, including controls for whether each building was rated by LEED or
Energy Star. Following the standard specification found in the hedonic pricing literature for
real estate assets (see ?), the natural log of prices, p,
5
is characterized by the set of all its
physical attributes, found in the vector, X
i
, including an indicator variable for the existence
of a green rating, x
green
, such that
p
i
= β
0
+ β
i
X
i
+ β
green
x
green
+
i
, (3)
5
The semi-log specification is used to correct for the skew in the distribution of office building prices.
5
where the βs are coefficients to be estimated and
i
is a building-specific residual. It is further
assumed that the preferences of the commercial real estate investors are solely determined
by the corresponding vector of attributes, including the ratings, that define the building. In
contrast to Equation (2), the hedonic specification found in these studies does not control
for the operating costs of the buildings nor does it control for the expected metropolitan-
level energy costs for the major fuels used by commercial office buildings; natural gas and
electricity. Thus, the introduction of an indicator variable for “green” building ratings is
likely to primarily account for the benefit stream associated with the ratings. The benefit
stream would be expected to have a positive effect (i.e., positive β
green
) although the effect on
prices might be statistically insignificant. The causal determinants of this benefit, however,
would be indeterminant. It could either be associated with real energy efficiency of the
building, although this is unmeasured in this specification, or it could be due to the “plaque-
in-the-lobby-effect” or other labeling related attribute effects (see, for example, the label of
an “architect designed building,” as in ?).
A primary focus of this research is to assemble a data set suitable for the empirical
estimation of Equation (2). Since Equation (2) measures current values as a function of
forward measures of fundamentals, our data requirements include forward measures for the
metropolitan-level net-income growth rates, as identified by Plazzi et al. (2010), as well as
building-level measures of the dynamics of key forward-looking factor input prices. Any em-
pirical analysis of the highly heterogeneous stock of commercial office properties must also
include numerous controls for the physical and utilization characteristics of these buildings,
as is typically done in the estimation of hedonic price estimates. Another focus of our empir-
ical work is to consider the affect of each building’s contractual leasing structure on observed
total operating expenses, Energy Star ratings, and asset values. Since the contractual struc-
tures of leases stipulate the way in which utility costs (primarily gas, electricity, water and
taxes) are allocated to tenants and the degree of control tenants have on these costs, we
develop measures for these leasing structures and consider several different empirical speci-
fications for the factors that are associated with the adoption of one leasing structure versus
another. Finally, we consider the relationship between the capital structure of commercial
office buildings and their energy risk characteristics.
3 The CoStar Data
CoStar Group data used in our analysis comprises two separate components: the Properties
data and the Comparable Transactions data. The Properties data includes information
on subletting, direct, or relet space that is currently available for a large sample of office
6
buildings in the United States. CoStar reports the “Weighted Average Rent,” if there is
rentable space available in the building; otherwise the Weighted Average Rent measure
appears as missing. The Weighted Average Rent is measured as the weighted average “asking
rent” for the available sublet, direct, or relet office space in each building at the time the
data are downloaded (in our case, August/September 2010). The property data file also
includes detailed information concerning the characteristics of the space that is available,
including the location of the building, the dominant lease contract structure in the building,
the amount of leaseable square footage available at the download date, an undated indicator
variable for whether the building has been Energy Star rated at any time between 1999 and
2010, and the most recent sale transaction information (date of sale and sales price) if there
have been recent transactions. There is also a large amount of brokerage contact information
since these data are intended for use by leasing brokers and tenants seeking to dispose of, or
obtain, office space.
There are a number of problems with the use of the CoStar measure of Weighted Average
Rent in a statistical analysis of the correlation of building attributes, such as indicators for
energy efficiency, and the market values of commercial office buildings. First, since there
can be no assurance that the CoStar quoted “asking rent” is ever achieved in any future
lease transaction, these rents cannot be viewed as directly equivalent to market prices. At
best these asking rents might be viewed as a noisy measure of the landlord’s evaluation of
the market value of the available space given the available space characteristics. A second
problem is that the CoStar measure of the Weighted Average Rent does not correspond to
a homogeneous combination of available rental space. Since sublet, relet, and direct space
would be expected to have very different quoted and realized rents per square foot, the
Weighted Average Rent cannot be readily compared across buildings. For these rents to be
comparable, additional information on the amount of each type of available space and the
rents for each would be required. Unfortunately, the asking rent per square foot for each type
of rental space is not reported in Costar. A final important limitation is sample-selection bias.
Since the Weighted Average Rent appears as a missing value for all buildings that are fully
leased, more poorly functioning buildings are likely to be systematically over-represented in
the CoStar data.
Given these problems with using the CoStar Weighted Average Rent measure as a proxy
for market value, we instead use the Costar Comparable Transaction data. As previously
discussed, we limit our analysis to comparable arms-length and confirmed market transac-
tions.
6
These data again have no information on the actual rents paid by current lessors
6
We eliminate all transactions for which there was a “non-arms-length” condition of sale due to such
factors as a 1031 Exchange, a foreclosure, a sale between related entities, a title transfer, among other
7
but they do have information on the actual confirmed transaction price and sale/recording
dates for each office building. The data also include information on the overall building
characteristics (building and lot square footage, typical floor area square footage, numbers
of floors, etc), how many tenants, the location, and quality characteristics of the building,
information on the first and second lien amounts, and the lien periodic payment amounts.
For a subset of these data, there is also information on the annual net operating income at
sale, the market-capitalization rate at sale, and the total annual expenses at sale.
Since the Comparables, or Transaction data, represent a subset of the Properties data,
or asking-rent data, in CoStar, we merge the two data sets together by building name and
address to obtain leasing characteristic information for the subset of office buildings that are
in both data sets. The merged set of Transactions and Property data included 15,133 office
buildings with complete records on all important covariates such as building square footage
or the number of tenants. We also analyze a smaller data set that includes information on
the leasing structure, annual net operating income at sale, the actual cap rate at sale, and
the total annual expenses at sale.
3.1 Energy Star Matching
The CoStar sample of 15,133 buildings was then merged with a data set obtained from U.S.
Environmental Protection Agency (EPA) that included all the buildings in the U.S. that
have obtained an Energy Star rating.
7
The Energy Star rating program was designed by the
EPA and the U.S. Department of Energy to promote energy efficiency in the U.S. commercial
real estate sector and thereby reduce greenhouse gas emissions. The program was started
in 1999. The Energy Star rating is based on comparative national data, obtained from the
Commercial Building Energy Consumption Survey, which set the annual benchmarks for
energy usage levels across property types. A building’s energy efficiency is measured as the
residual between the actual and predicted energy usage of the building using actual utility
bills. To receive an Energy Star label, a building must score in the top quartile of the EPA’s
energy performance rating system and must meet designated indoor air quality standards.
For each building, we obtained information on when the Energy Star rating was obtained
conditions. All of these sale conditions would affect prices due to the trading of tax basis in the case of 1031
exchanges or the auction structure in the case of foreclosure. Instead, we focus only on market transactions
between unrelated persons.
7
Many buildings in this sample were Energy Star rated multiple times and these ratings are often non-
monotonic in time (sometimes lower ratings are obtained at later dates). This non-monotonicity may arise
because the Energy Star rating is relative to the population mean performance of office buildings. Thus,
if an office building simply maintained its energy consumption profile, its ranking might fall if the overall
population of U.S. buildings increases its energy efficiency.
8
and the level of the rating the building received.
8
The merged data sets led to 545 matches
for Energy Star rated buildings. However, many of these matches were for Energy Star
ratings that post-dated the actual observed transaction date.
9
Using data on the actual date
of each Energy Star grant, we matched 141 buildings that had an Energy Star rating by the
time of sale.
3.2 Local Weather Data Matching
The weather data were obtained from Wolfram Schlenker at Columbia University.
10
The
weather station data are based on a rectangular grid system, called PRISM, that was devel-
oped at Oregon State University and covers the contiguous United States.
11
The weather
data include 471,159 grid points representing 2.5 mile squares with non-missing data. The
data include the minimum and maximum temperature (Celsius) and the total precipitation
(cm) for each day of a year for all of the 471,159 grids in the United States from 1950 through
2010. These data are interpolated from PRISMs monthly weather station averages to daily
data and we aggregate them back into monthly data for our analysis. We associate the
past twelve months of weather data for each building in the CoStar data with the weather
data associated with the nearest grid point in the Schlenker, et al. data. Further details
concerning the structure of the weather data are reported in the Appendix.
3.3 Energy Auction Data Matching
Incorporating information on the energy factor inputs for U.S. office building exposure re-
quires a careful accounting for the institutional and contractual details of the regional and
sub-regional gas and electricity markets in the U.S. We use data purchased from Platts (the
data vendor) and compute a daily forward curve for power purchased on-peak and off-peak for
all the trading hubs represented in our metropolitan areas.
12
Platts gathers information on
the power forward market from active brokers and traders and through the non-commercial
departments (back offices) of companies. Since October 2007, this information has been
augmented with auction prices from the Intercontinental Exchange (ICE) to form Platts for-
ward market power daily assessment. Since more liquid locations and shorter term packages
8
These rating vary between a minimum of 75 and a maximum of 100.
9
Costar does not account for the date the Energy Star rating was received.
10
http://www.columbia.edu/
~
ws2162/
11
http://www.prism.oregonstate.edu/
12
The location of these hubs is detailed in the Appendix. The subset of hubs that is described in the
Appendix is currently being expanded and will be included in the next version of the paper. We currently
use geographic interpolation to obtain estimates for the hubs for which we do not currently have the auction
data. The additional five hubs will be included in the next version of the paper.
9
trade more frequently on ICE, while less liquid locations and longer term packages trade
more frequently over-the-counter (OTC), Platts is able to combine these sources to build a
comprehensive picture of the forward market. Details of the methodology are described in
the Appendix.
The raw data from Platts was formatted with single entries for each forward package.
For a given trading date, a power hub, and a type of contract - on peak and off-peak -
there are single entries for the mark-to-market price for each forward package. This scheme
characterizes the term-structure of power prices for a given trading date for contracts of
varying maturities. Since the hub markets are defined geographically we then develop two
measures for each building: 1) the 1-12 month daily average forward price per month (a
measure of the short term contract forward price) that is measured contemporaneously, with
a six-month lag and with a twelve-month lag; 2) the shape of the forward curve measured
as the difference between the daily average for 1-12 month and 25-36 month contracts,
standardized by the number of months in the curve, that is also measured contemporaneously,
with a six-month lag and with a twelve-month lag. These measures were then matched to
each building according to the electricity forward market hub that serves the building’s
location and were matched to the observed the month of the building’s sale date. The
electricity prices are quoted as $/MW h (Mega Watts x hour) and Platts publishes these
prices as of the delivery, or flow date, of the contract. Our use of delivery prices justifies our
contemporaneous merges between electricity forward delivery price and the contemporaneous
sale date of the building.
The resource costs (wholesale) price dynamics for natural gas are measured similarly to
those of the electricity hubs. One important difference, is the natural gas market is bench-
marked to a single auction at the Henry Hub. Following our strategy for the electricity
prices and slopes, we measure the 1-12 month forward prices and the slopes for the Henry
Hub. After the deregulation of the wholesale market for natural gas in the mid 1990’s, the
New York Mercantile Exchange (NYMEX) launched trading for monthly futures contracts
with similar characteristics to those of crude oil. The standard NYMEX natural gas futures
contract specifies physical delivery of 10,000 MMBtu (millions of British thermal unit) rat-
ably delivered into Henry Hub - Louisiana. Until early 2000’s NYMEX provided monthly
contracts covering maturities of about 36 months out. After that the range of maturities was
extended and it currently covers more then six years (72 months) out on a monthly basis.
The NYMEX website provides more details on how the contracts are traded and the rules
for settlement.
There is also an extensive network of natural gas pipelines connecting the production
basins to large consumption areas (mainly large populated urban centers) and wholesale
10
physical natural gas trading occurs in different hubs distributed in the continental U.S.
These hubs are key points in the pipeline grid characterized by either being interconnections
between major pipelines and/or access points to public utility gas companies. Of all those
hubs, Henry Hub is the benchmark for price quotation. Henry Hub’s importance comes
from its location as an interconnecting point for multiple pipelines and because it is the
most liquid hub for trading spot and futures contracts. Prices for other hubs (spot and OTC
forwards) are typically quoted as a basis to Henry Hub. These basis quotes are, most of the
time, a very small fraction of the full benchmark quote. We follow the market conventions
and compute the near natural gas price as the Henry Hub monthly average of daily 1-12
month forward prices and measure the slope as the difference between the near price and
the 60-72 month forward prices. We again compute these value contemporaneously, with a
six-month lag and with a 12-month lag for each date. We then merge these time series data
to the date of the observed sales transactions for each office building. Other specifics of our
natural gas measurement is described in the Appendix.
3.4 Summary Statistics for the CoStar Transaction Data
As previously discussed, we focus on the forty- three metropolitan “office” market areas
that account for the highest levels of employment in the category of Information, Finance,
and Professional and Business Services, as measured by the Bureau of Labor Statistics in
2010.
13
We represent the office market location of the building using market area designations
developed by CoStar. In Table 2, we report the frequency of office building arms-length
transactions that occurred from 2001 through 2010 for which we have complete price and
characteristic information. As shown in the table, we have good transaction coverage for
all of the large U.S. office markets identified by the Bureau of Labor Statistics. Our total
sample size is 15,133 office properties.
Table 3 presents the distribution of transaction dates of sale for buildings that traded in
our sample. The heavy trading volumes in the years 2005 through 2007 reflect the growth
of the Commercial Mortgage Backed Securities market (an important source of mortgage
lending for office transactions) during this period and the transaction boom fueled by the
availability of cheap credit.
Table 4 presents the distribution of office buildings in the sample that had obtained at
least one energy star rating by 2010. As shown, we were able to match to 547 Energy Star
rated buildings in the U.S. EPA Energy Star ratings reports. When we further narrow the
definition of the Energy Star rated buildings to buildings that were Energy Star rated at the
13
Bureau of Labor Statistics, Employment Hours and Earnings, State and Metro Area, http://www.bls.
gov/sae/data.htm.
11
Table 2: Market Location of CoStar Transaction Data
This table presents the sample of office buildings that traded in arms-length transactions between
2001 and 2010 for which we have sales price, sales date, and information on the square footage of
the building. The data were obtained from the COSTAR transactions data base.
Costar Market Area Number of Sales Percentage of Total Cumulative Frequency Cumulative Percentage
Atlanta 915 6.05 915 6.05
Austin 93 0.61 1008 6.66
Baltimore 370 2.44 1378 9.11
Boston 510 3.37 1888 12.48
Charlotte 102 0.67 1990 13.15
Chicago 909 6.01 2899 19.16
Cincinnati/Dayton 156 1.03 3055 20.19
Cleveland 154 1.02 3209 21.21
Dallas/Fort Worth 306 2.02 3515 23.23
Denver 617 4.08 4132 27.3
Detroit 166 1.1 4298 28.4
East Bay/Oakland 188 1.24 4486 29.64
Hartford 48 0.32 4534 29.96
Houston 218 1.44 4752 31.4
Indianapolis 47 0.31 4799 31.71
Inland Empire (California) 380 2.51 5179 34.22
Kansas City 161 1.06 5340 35.29
Las Vegas 528 3.49 5868 38.78
Long Island (New York) 283 1.87 6151 40.65
Los Angeles 799 5.28 6950 45.93
Marin/Sonoma 33 0.22 6983 46.14
Milwaukee/Madison 30 0.2 7013 46.34
Minneapolis/St Paul 184 1.22 7197 47.56
Nashville 52 0.34 7249 47.9
New York City 243 1.61 7492 49.51
Northern New Jersey 637 4.21 8129 53.72
Orange (California) 464 3.07 8593 56.78
Orlando 323 2.13 8916 58.92
Philadelphia 762 5.04 9678 63.95
Phoenix 1351 8.93 11029 72.88
Pittsburgh 92 0.61 11121 73.49
Portland 179 1.18 11300 74.67
Sacramento 307 2.03 11607 76.7
San Antonio 47 0.31 11654 77.01
San Diego 292 1.93 11946 78.94
San Francisco 185 1.22 12131 80.16
Seattle/Puget Sound 410 2.71 12541 82.87
South Bay/San Jose 161 1.06 12702 83.94
South Florida 707 4.67 13409 88.61
St. Louis 119 0.79 13528 89.39
Tampa/St Petersburg 527 3.48 14055 92.88
Washington DC 1022 6.75 15077 99.63
Westchester/So Connecticut 56 0.37 15133 100
12
Table 3: Transaction Dates for the Arms-Length Office Building COSTAR Transactions
This table presents the frequency of trades for the office buildings in our sample. Our sample of
buildings are office properties that traded in arms-length transactions between 2001 and 2010 for
which we have sales price, sales date, and information on the square footage of the building. The
data were obtained from the CoStar transactions data base.
Sale Year Number of Sales Percentage of Total Cumulative Frequency Cumulative Percentage
2001 10 0.07 10 0.07
2002 227 1.5 237 1.57
2003 1806 11.93 2043 13.5
2004 2457 16.24 4500 29.74
2005 2534 16.74 7034 46.48
2006 2570 16.98 9604 63.46
2007 2641 17.45 12245 80.92
2008 1757 11.61 14002 92.53
2009 935 6.18 14937 98.7
2010 196 1.3 15133 100
time of their sale, we have only 141 Energy Star Ratings at the time of sale. This feature
of the data arises because the incidence of Energy Star rated buildings has been growing.
However, most buildings rated by Energy Star received their ratings at the end of the sample
in 2008, 2009 and 2010, and typically these rating dates are several years after the properties
actually sold.
In the Appendix, we map the location of the Energy Star rated buildings in the Los
Angeles and San Franciso markets. The maps highlight the important geographic structure
of the Energy Star rated building locations. For the most part, the Energy Star rated
office buildings are located in more central locations, the Central Business District or the
Sub-Regional Business District, within the CoStar Markets. A further geographic detail,
which is not obvious from Table 4, is that the preponderance of these Energy Star certified
buildings are located in the states of California, Florida, Texas, New York/New Jersey,
and Washington DC/Maryland. Several markets have only one or two Energy Star rated
buildings.
Table 5 provides the summary statistics for other important characteristics of the office
building transaction data. In the top panel of the table, we report summary statistics for
the 1-12 month power and natural gas forward prices that were observed at the time the
building traded. We also report the values for the slope of the forward curve for power
and natural gas on the sale date. As reported, the average electricity forward price (in
$/MWh), that is observed at the sale date for the traded office buildings is $68.65/MWh,
the standard deviation is $19.23/MWh, and there is considerable variability, with a high of
13
Table 4: Energy Star Ratings Frequencies between 2001 and 2010
The upper panel of the table presents the numbers of office buildings that even received at least
one Energy Star rating between 2001 and 2010. The lower panel of the table reports the number
of office buildings in the sample that had an Energy Star rating by the time of their sale.
Energy Star Number Percentage Cumulative Cumulative
Status of Sales of Total Frequency Percentage
Never EnergyStar Rated in period 14586 96.39 14586 96.39
EnergyStar Rated in period 547 3.61 15133 100
Never EnergyStar Rated by time of Sale 14992 99.07 14992 99.07
EnergyStar Rated by time of Sale 141 0.93 15133 100
Total Energystar Annual Ratings (1999 - 2010) 1222
$161.71/MWh and a low of $30.78/MWh. Although not shown, there is also considerable
variability in these prices across hub regions, so office buildings in different hubs at the same
time period experienced different prices. The average slope of the power forward curve is
slightly downward sloping per month $-0.819. However, here again there is considerable
time-series and hub variation, with observations with very steep forward curves indicating
expectations that power prices/MWh were expected to rise, $13.67/MWh/month, and very
steeply downward sloping forward price curves, $-21.74/MWh/month.
As previously discussed, the natural gas forward prices only vary in the time series since
we measure all buildings at the Henry Hub benchmark forward price following industry
convention. As shown in Table 5, the average forward price (in $/MMBtu) is $7.38 with a
standard deviation of $2.047. The maximum price was less that twice the average price. The
observed slope of the forward curve at the transaction date was slightly negative at $ -0.18
MMBtu/month and, similar to the power markets, the minimum and maximum values vary
between positively and negatively sloped forward price curves.
The third panel of Table 5 provides the summary statistics for the twelve months prior
to the sale date of the property. We report the average standard deviation of weather data
for the minimum temperature, maximum temperature, and the precipitation over the prior
twelve months for each building. As shown, the average standard deviation for the maximum
temperatures was 69.14 degrees Farenheit and was 58.49 degrees Farenheit for the minimum
temperatures. The precipitation standard deviations are significantly smaller at .045.
In the third panel of Table 5, we report the summary statistics for the trading price
and the building characteristics of the arms-length transactions. The average observed price
per square foot was $183.89, with a standard deviation of $108.72 per square foot, and
the average building size was 54,747.95 square feet with a standard deviation of 122,254.48
14
Table 5: Summary Statistics for Office Buildings Sold in Arms-Length Transactions between 2001 and 2010
This table presents the summary statistics for buildings that sold in arm-length-transactions between 2001 and 2010. The Transaction
data were obtained from the COSTAR Transactions data base. The energy factor price data were obtained from Platts. The weather
and precipitation data were obtained from Wolfram Schlenker
Variable N Mean Std. Dev. Minimum Maximum
Average price of the 1-12 month power
forward contract by transaction hub ($/M W h) 15133 68.652 19.237 30.778 161.717
Slope of the power forward curve by transaction hub ($/MW h/month 15133 -0.819 3.172 -21.744 13.671
Average price of the 1-12 month natural gas
forward contract - Henry Hub ($/MM Btu) 15133 7.381 2.047 2.501 12.721
Slope of the natural gas forward curve - Henry Hub ($M MBtu/month) 15133 -0.181 0.486 -1.657 1.005
Standard deviation of the maximum temperature (Farenheit) 15133 69.135 39.506 2.223 161.269
Standard deviation of the minimum temperature (Farenheit) 15133 58.493 32.235 1.914 133.617
Standard deviation of preciptation (Centimeters) 15133 0.045 0.039 0.000 0.366
Transaction price ($/sqft) 15133 183.889 108.729 20.000 798.460
Number of floors 15133 3.437 5.201 1.000 110.000
Age of the Building (years) 15133 30.784 28.616 1.000 198.000
Typical floor area square footage 15133 14058.730 18890.540 1000.000 455304.000
Building size (Square Feet) 15133 54747.950 122254.480 360.000 3781045.000
Indicator Variable for Multi-Tenant Building 15133 0.436 0.496 0.000 1.000
Indicator variable for class A buildings 15133 0.123 0.328 0.000 1.000
Indicator variable for renovation prior to year of sale 15133 0.013 0.114 0.000 1.000
Total expenses per square foot at year of sale ($/sqft) 1470 6.843 4.083 2.040 73.105
Net operating income per square foot at year of sale ($/sqft) 1527 11.910 6.977 4.034 71.489
Market capitalization rate at year of sale (%) 2322 7.733 1.563 2.800 13.140
15
square feet. The average number of floors was 3.44 and the largest building had 110 floors.
The typical rented square footage per tenant was 14,058.73 square feet in these buildings and
the standard deviation was 18,890.54. Forty four percent of the buildings in the transaction
data are multi-tenant buildings and twelve percent of the buildings are class A buildings.
Only one 1.3% of the buildings in the sample were renovated prior to the sale.
For a subset of the transaction data, we have information on the net operating income,
the total operating expenses, and the observed capitalization rate at the time of the sale.
14
As reported in the bottom panel of Table 5, for the sub-sample of sales, the average net
operating income per square foot at the time of sale was $11.91 and the standard deviation
was $6.98. The total expenses per square foot at the time of sale was $6.84 and the standard
deviation was $4.08. Finally, the observed capitalization rate at the time of sale was 7.73%
with a minimum value of 2.8% and 13.14% over the period 2001 to 2010.
4 The Investor’s Energy Star Rating Decision
We first consider the exogenous factors that may lead a building owner to apply for and
successfully achieve an Energy Star rating for a commercial office building. As previously
discussed, applying for an Energy Star rating involves significant costs in the form of the
time and effort required for engineers to verify utility bills and certify the level of air quality
in the building. There may also be costs associated with retrofitting the building, if the
current operating performance is below the top quartile of performance. Following the logic
of standard investment rules, we would expect the investor to undertake the Energy Star
application process and, if need be, retrofit the building to successfully achieve an Energy
Star rating, if the present value of the benefits from that investment exceeded its costs.
15
As
is clear from the previous discussion of the asset price equation, Equation (2), the increment
to the asset price could arise from the increase in gross income, a decrease in vacancy rates,
a decrease in total operating costs, an increase in the net operating income growth rate, or
a decrease in the risk of the asset due to reductions in the volatility of the buildings cash
flows. Whatever the source of the benefits from the rating, however, the increment to the
asset’s value after the successful Energy Star rating must exceed the cost of obtaining the
rating given the investor’s opportunity costs.
An additional dilemma for the econometrician is that the underlying fundamental fac-
14
The capitalization rate is the discount rate that translates the observed net operating income into the ob-
served transaction price at the time of sale, assuming an infinite investment horizon,
P
∞
t=1
NOI
capitalization rate
=
Sales P rice.
15
Of course, real options considerations might also enter this calculus leading to the consideration of the
second moments of fundamental factors. Grenadier (2005)
16
tors defined in Equation (2) are unobserved. Instead, the econometrician observes only an
outcome variable that is one if the decision to obtain an Energy Star rating meets the invest-
ment threshold for the investor. Future observations on the price of the asset are, of course,
conditioned on the prior decision to obtain, or not to obtain, the Energy Star rating and
this decisions could affect observed levels and dynamics of asset prices in important ways.
A further problem, is that the Energy Star decision is often a dynamic contracting problem
that is solved contemporaneously with leasing and debt contracting decisions. This further
complication implies that the Energy Star decision may be co-determined with these other
contracting decisions.
In its simplest form, the utility an investor derives from obtaining an Energy Star rating
in a given period can be associated with a linear function of building characteristics and
other exogenous market variables, v
it
, affecting costs and building value,
U
it
= γv
it
+ µ
it
, (4)
where µ
it
is a residual. Utility is a latent variable that is not observed, however, we do
observe the choice made by the investor at each period. Obtaining an Energy Star rating
thus corresponds to a response variable, y
it
, with a value to one. If the Energy Star rating
is achieved, this implies that the latent utility is positive,
y
it
= 1 ⇒ U
it
> 0. (5)
Under the assumption that the residual follows an i.i.d extreme value distribution, the
probability of become Energy Star rating is given by the logistic function.
P r(y
it
= 1) = P r(γv
it
+ µ
it
> 0) = P r(µ
it
> −γv
it
) =
1
(1 + exp(γv
it
)
. (6)
The coefficients of this choice behavior can be estimated using maximum likelihood. The
likelihood function is calculated by aggregating the probability of the observed choice stream
for each building.
logL(θ; v) = Σ
N
n=1
ln[P r(y
it
: y
nt
= 1 ∀t = 1, . . . , T
n
)]. (7)
We report the results of the maximum likelihood estimation for the Energy Star choice
in Table 6. Since we only observe the year of the Energy Star rating, we compute the annu-
alized average near contract prices and the annualized value of the slope for power and gas
for each building. We also have measures for the quality of the building, the age and the
17
square footage at the Energy Star grant date and the quality level of the building. Many of
the Energy Star rated buildings in the sample are rated multiple times over the period 1999
through 2010. For every building in the sample, we construct a panel of annual observations
on market and building characteristics including an indicator variable for whether the build-
ing became Energy Star rated in that year. Our estimator accounts for the path dependence
of these decisions since a sub-sample of the buildings were Energy Star rated every year, some
buildings were Energy Star rated more than once but not every year, some were Energy Star
rated only once in the sample period, and some buildings were never Energy Star rated over
the period. Since the Costar data only gives us a cross-sectional snapshot of the building at
the transaction data we only use building characteristics that are measurable over time, such
as age, building square footage, and quality level. We merge the building level characteris-
tics with time varying market characteristics for the energy forward prices and the weather
variables and these are measured year-by-year from the building’s transaction date.
16
As shown in Table 6, the decision to Energy Star rate a building appears to be importantly
related to the energy factor prices in markets. There is a statistically and economically
negative relationship between the six month lagged price level of the one to twelve month
electricity contract and a statistically positive relationship with the slope of the electricity
forward curve indicating expectations are for electricity prices to rise. The interpretation of
these results is that a one unit change in the average level of local hub electricity prices would
be expected to lead to a -1.8% decrease in the log odds ratio that the Energy Star rating
is achieved, whereas a similar magnitude change in the expectation that the hub price for
electricity delivered in five years would lead to a 3.47% increase in the odds that an Energy
Star rating would be achieved, holding other effects constant. Changes in the level of Henry
Hub natural gas prices do not have a statistically significant effect on the likelihood that an
Energy Star rating is achieved, whereas expectations that natural gas prices are expected to
rise have a statistically negative effect on the probability of achieving an Energy Star rating.
MSAs with higher variance in minimum temperature are shown to be statistically positively
related to the decision to obtain an Energy Star rating in a given year, whereas increased
variance in maximum temperatures is negatively associated with obtaining an Energy Star
rating. These results suggest that it is the variance in heating that is the primary driver of
Energy Star rating decisions. As expected, the results show that Class A buildings are more
likely to obtain an Energy Star rating as are larger buildings and newer buildings.
Of course, an important caveat with our Logit results is that we have no additional
controls for whether the buildings in the CoStar sample are actually energy efficient, whether
16
We were only able to obtain local weather data for each building through 2005. We use the 2005 weather
metrics for Energy Star rating events after 2005.
18
Table 6: Logit Estimation of the Energy Star Rating Choice
This table presents maximum likelihood estimates for the probability that a building sucessfully
received an Energy Star rating in a given year as a function of local energy-market and weather
characteristics and building and market characteristics. The annual Energy Star rating indicator
for each building was obtained from the U.S. Environmental Protection Agency, the office building
data was obtained from CoStar, the electricity forward curve data was estimated using data from
Platts, the natural gas data was obtained from NYMEX, and the weather data was obtained from
Wolfram Schlenker. The first two columns report the results for estimating the time path of Energy
Star ratings for building between 1999 and 2010 (true for 1222 annual Energy Star building ratings).
The reported test statistics are distributed Chi Squared.
Parameter Standard
Variable Estimate Error
Intercept -5.949
∗∗∗
0.317
Annual Average of the six month lag price of the 1-12 month power
forward contract at sale by transaction hub -0.018
∗∗∗
0.004
Annual Average of the six month lag slope of the power forward curve 0.0347
∗∗
0.016
Annual average of the six month lag price of the 1-12 month
natural gas forward contract at sale -0.035 0.026
Annual average of the six month lag slope of the natural gas forward curve -0.055
∗∗∗
0.0125
Standard Deviation of 12 Month Maximum Temperature -0.014
∗∗∗
0.005
Standard Deviation of 12 Month Minimum Temperature 0.018
∗∗∗
0.006
Standard Deviation of 12 Month Precipitation -0.158 1.133
Age of the Building -0.015
∗∗∗
0.002
Building square footage 0.020
∗∗∗
0.001
Indicator variable for class A buildings 2.847
∗∗∗
0.081
Market Fixed Effects Yes
Year Fixed Effects Yes
Likelihood Ratio Test 4184.83
∗∗∗
Number of Observations (Balanced panel building per year) 173,640
19
or not they have an Energy Star rating. Thus, our specification only represents decisions
whereby the labeling decision was an NPV positive decision, not whether overall objective
energy efficiency measures lead buildings, on average, to obtain an Energy Star rating. In the
next sections of the paper, we will further explore the operating expense and net operating
income characteristics of buildings to better understand the relationship between the Energy
Star label and relative building operating performance.
5 Empirical Results with Transactions Prices
As previously, discussed due to data limitation all previous studies of the effects of energy
metrics or environmental and architectural factors on commercial office building values have
applied hedonic approaches similar to that of Equation (3) (for examples, see Wheaton and
Torto, 1994; ?; ?; ?; ?; ?, among others). These studies usually find that the benefits of
energy certification or the environmental/architectural attributes of buildings are positive
factors leading to a premium on buildings with these attributes. These results, however, say
nothing about whether the investment in the certificate or the environmental or architectural
attributes are Net Present Value positive or whether, in the case of energy certification the
buildings are in effect energy saving. To answer the question whether energy certification,
based on quantifiable energy related metrics like the Energy Star ratings, requires a spec-
ification that is closer to Equation (2) and controls for costs and expected energy-related
factor input prices.
In Table 7, we report the results of two specifications: in the first four columns we report
the results for the hedonic specification; in the fifth and sixth columns we report the results
for the forward looking asset pricing specification defined by Equation (2). As shown in
the Table, the hedonic specification indicates that higher building values are associated with
higher near-term power forward prices and a more steeply sloped forward curve even after
controlling for market fixed effects. Thus, the level of net operating income or its growth
rate appears to more than compensate for exposure to increase electricity costs, which is the
primary energy exposure for commercial office buildings. Higher prices for natural gas near-
term futures contracts have a statistically significant and negative association with building
prices per square foot indicating that the cost effect of this factor input is not compensated
by rents. The effects of the standard deviation in the twelve-month maximum temperature
has a statistically significant negative effect of price and the and the standard deviation of
minimum weather temperature has a smaller positive effect on the price per square foot. As
shown, class A buildings and those buildings that were Energy Star rated prior to the sale
have economically and statistically significant positive associations with building values. The
20
Table 7: Regression Results for Office Buildings that Sold in 2001 through 2010
This table presents the regression results for buildings that sold in arms-length transactions between 2001 and 2010. The dependent
variable is the log of the price per square foot of the building at sale. These data were obtained from the COSTAR Transactions data
base. The Energy Star rating indicator for each building was obtained from the U.S. Environmental Protection Agency, the office building
data were obtained from CoStar, the electricity forward curve data was estimated using data from Platts, the natural gas forward curve
was estimated using data from NYMEX, and the weather data were obtained from Wolfram Schlenker.
Parameter Standard Parameter Standard Parameter Standard
Intercept 5.346
∗∗∗
0.109 5.567
∗∗∗
0.147 5.605
∗∗∗
0.202
Six month lag price of the 1-12 month
forward contract at sale by transaction hub 0.007
∗∗∗
0.001 0.004
∗∗
0.002 0.002 0.002
Six month lag slope of the power forward curve 0.015
∗∗∗
0.003 0.003 0.006 0.017
∗∗∗
0.007
Six month lag price of the 1-12 month
natural gas forward contract at sale -0.025
∗∗∗
0.006 -0.017
∗
0.010 -0.015 0.014
Six month lag slope of the natural gas forward curve -0.012 0.022 0.023 0.042 -0.121
∗
0.068
Standard Deviation of 12 Month Maximum Temperature -0.009
∗∗∗
0.001 -0.011
∗∗∗
0.002 0.008
∗∗∗
0.003
Standard Deviation of 12 Month Minimum Temperature 0.005
∗∗∗
0.002 0.009
∗∗∗
0.003 0.006
∗∗
0.003
Standard Deviation of 12 Month Precipitation 0.362 0.283 0.338 0.464 -0.478 0.536
Number of floors 0.002 0.002 0.000 0.003 0.001 0.003
Age of the Building -0.002
∗∗∗
0.000 -0.004
∗∗∗
0.001 -0.003
∗∗∗
0.001
Building square footage 0.002
∗∗
0.040 0.002
∗∗
0.001 0.001 0.001
Typical floor area square footage -0.046
∗∗∗
0.038 -0.016
∗∗∗
0.006 -0.001 0.008
Indicator variable for renovation prior to sale -0.084
∗∗∗
0.001 -0.115
∗∗∗
0.024 -0.043 0.039
Indicator variable for class A buildings 0.210
∗∗∗
0.004 0.207
∗∗∗
0.022 0.182
∗∗∗
0.035
Indicator Variable for Multi-Tenant Building -0.056
∗∗∗
0.012 0.057 0.077 0.110 0.121
Market capitalization rate at year of sale -0.104
∗∗∗
0.010
Net operating income per square foot at year of sale 0.461
∗∗∗
0.042
Total expenses per square foot at year of sale -0.193
∗∗∗
0.041
Indicator variable for Energy Star rating by time of sale 0.134
∗∗∗
0.019 0.097 0.067 0.079 0.089
Market Fixed Effects Yes Yes Yes
Year Fixed Effects Yes Yes Yes
Clustered Standard Errors (Sale date and Market) Yes Yes Yes
R
2
0.306 0.409 0.599
Number of Observations 15133 1470 1470
21
Energy Star result is similar to results from hedonic specifications in prior papers. Building
attributes that negatively affect building values include age, multi-tenant buildings, and
large rentable floor plates.
In the third and fourth columns of Table 7, we re-estimate the hedonic regression for
a sub-sample of buildings for which we have data required to estimate the forward looking
asset pricing specification defined by Equation (2). As shown, the results are nearly identical
to those for the larger sample although statistical significance is lessened due to the smaller
sample size. The smaller sample reduces statistical power and now the Energy Star rating,
which is based on quantitative measures of energy efficiency of commercial office buildings,
does not have a statistically significant positive association with the transaction prices of
these buildings. Without further controls, it is not possible to conclude that these effects
are due to the relative efficiency of these buildings or to the “plaque-on-the-wall” effect of
the certification.
In the fifth and sixth columns of Table 7, we report the results where we control for
the total expenses, market capitalization rate capturing the expected growth rate in rents,
and the net operating income for the building at the time of sale. In this specification
we find that the slope of the electricity forward curve is positively associated with log of
building price per square foot and the slope of the natural gas forward curve is positively
associated with log prices suggesting that auction market bets that energy prices will rise
are associated with higher expenses per square foot. The standard deviation of minimum
and maximum temperatures also has a statistically significant and positive effect on the log
of prices. These results appear to suggest that the primary driver of higher weather related
costs are associated with air conditioning costs in the summer rather than heating costs
in the winter months. Total expenses, the capitalization rate, and net operating income
are all economically and statistically significant with the anticipated signs. Buildings with
higher total expenses and that were exposed to higher market capitalization rates had lower
transaction prices whereas everything else equal, buildings with higher net incomes have
higher prices. The revealing result is that controlling for operating expenses, factor prices,
and interest rates eliminates the statistically significant and positive Energy Star channel
to transaction prices. This result suggests that the Energy Star rating appears to have a
lesser effect on transaction prices that is not statistically significantly different from zero
once appropriate controls for expenses, rents, and interests are introduced. Of course, our
measure of operating expenses includes property taxes, labor costs, and utilities (primarily
electricity, gas, and water), so it might be highly correlated with the Energy Star metric
since utilities are on average about 20% to 30% of total operating expenses nationally.
17
17
Computed by the authors using various BOMA publications.
22
To better understand the relationship between the Energy Star rating and total operating
expenses, net operating income, and the capitalization rates at sale of the office buildings,
we report three separate regressions: i) net operating income per square foot; ii) operating
expenses per square foot; and iii) the capitalization rate. The results of these regressions
are reported in Table 8. Overall as shown, the Energy Star rating of the building does
not have a statistically significant effect on operating expenses, net operating income, or
the capitalization rate of transacted buildings. In contrast, the higher values of the six
month lag in the one to twelve month electricity forward prices have a statistically significant
positive association with net operating expenses, the slope of the electricity forward curve
is associated with higher total operating expenses, and the level of the Henry Hub natural
gas forward prices is positively associated with higher operating expenses, as expected. The
variance in the minimum weather temperature conditions over the twelve months prior to the
sale also has a statistically significant positive effect on operating expenses. More variance
in the maximum temperature appears to be associated with lower net operating income
and higher market capitalization rates, and does not have a statistically significant affect
on operating expenses. There appear to be scale economies associated with larger buildings
and larger rentable floor plates, however, class A buildings and multi-tenant buildings have
economically and statistically significantly higher total operating expenses.
Given the results in Table 8, the net operating income and operating expenses appear
themselves be functions of the local supply and demand factors, suggesting that controls
should be introduced for the joint-endogeneity of net operating income and operating ex-
penses in a building prices specification for Equation (2). Since the market capitalization
rate, is in large part a function of the interest rate, we assume that it is exogenous. In
Table 9, we report the results for a three stage least squares estimation of endogenous net
operating income and operating expenses as a function of exogenous factors such as the local
energy factor input prices for natural gas and electricity, the local weather conditions, build-
ing characteristics, local market fixed effects, and year fixed effects and the Equation (2)
specification with the instrumented value of net operating income, the instrumented value of
operating expenses, and the market determined capitalization rate. Again, the three stage
least squares estimation is using the sub-sample of transactions for which we have infor-
mation on operating expenses, net operating income, and the market capitalization rate at
sale. As shown in Table 9, operating expenses per square foot have a statistically signif-
icant and positive association with the slope of the natural gas forward curves, implying
that the higher the future bets from the auction markets on the cost of this factor input
the higher the expected operating costs of the building. The energy factor inputs appear
to have no statistically significant effect on operating income. More variance in the max-
23
Table 8: Regression Results for Net Operating Income per Square Foot, Capitalization Rate at Sale, Total Operating Expenses
Per Square Foot for Office Buildings that Sold 2001 through 2010
This table presents the regression results for regressions of the net operating income per square foot, the capitalization rate at sales,
and the total operating expenses per square foot of office buildings that sold in arm-length-transactions between 2001 and 2010 on
energy costs, temperature variances, building characteristics, and whether the building was Energy Star before the sale. These data were
obtained from the COSTAR Transactions data base. The energy forward curve data was estimated using data from Platts and NYMEX
and the weather data was obtained from Wolfram Schlenker.
Log Net Operating Market Capitalization Log Operating
Income per Sq. Ft. Rate Expenses per Sq. Ft.
Parameter Standard Parameter Standard Parameter Standard
Variable Estimate Error Parameter Error Parameter Error
Intercept 2.553
∗∗∗
0.154 9.33
∗∗∗
0.380 1.161
∗∗∗
0.175
Six month lag price of the 1-12 month
forward contract at sale by transaction hub 0.003 0.002 -0.010 0.006 0.004
∗∗
0.002
Six month lag slope of the power forward curve 0.003 0.008 0.005 0.019 0.015
∗
0.006
Six month lag price of the 1-12 month
natural gas forward contract at sale 0.000 0.012 -0.023 0.024 0.022
∗∗
0.010
Six month lag slope of the natural gas forward curve 0.086 0.063 -0.219 0.137 -0.033 0.047
Standard Deviation of 12 Month Maximum Temperature -0.006
∗∗∗
0.002 0.012
∗∗
0.005 -0.003 0.003
Standard Deviation of 12 Month Minimum Temperature 0.004 0.003 -0.013
∗∗
0.005 0.006
∗∗
0.003
Standard Deviation of 12 Month Precipitation -0.055 0.469 0.821 0.957 1.118 0.528
Number of floors 0.000 0.003 -0.030
∗∗∗
0.009 0.016
∗∗∗
0.003
Age of the Building -0.001
∗∗
0.000 -0.003
∗
0.002 -0.001 0.001
Building square footage 0.001 0.001 0.002 0.003 -0.003
∗∗∗
0.001
Typical floor area square footage -0.020
∗∗∗
0.005 0.012 0.017 -0.011
∗
0.005
Indicator variable for renovation prior to sale 0.024 0.031 -0.116 0.091 -0.091 0.072
Indicator variable for class A buildings 0.035 0.031 -0.370
∗∗∗
0.081 0.110
∗∗∗
0.035
Indicator Variable for Multi-Tenant Building -0.092
∗∗
0.033 0.021 0.017 0.216
∗∗∗
0.032
Indicator variable for Energy Star rating by time of sale 0.004 0.075 -0.096 0.203 -0.076 0.066
Market Fixed Effects Yes Yes Yes
Year Fixed Effects Yes Yes Yes
Clustered Standard Errors (Sale date and Market) Yes Yes Yes
R
2
0.282 0.349 0.313
Number of Observations 1470 1470 1470
24
imum temperature in the local markets has a statistically significant negative association
with operating income per square foot and more variance in precipitation has a statistically
significant positive effect on operating costs after controlling for local market fixed effects.
More floors and multi-tenant buildings are associated with statistically significantly higher
operating expenses and larger floor plates rented by tenants are associated with lower net
operating income probably due to the discounts afforded to larger block rentals.
Instrumenting for operating expenses and net operating income in the price equation
leads to results that are very similar to those reported in Table 7 that does not use a Three
Stage Least Squared Estimator. As expected from Equation (2), the capitalization rate, the
net operating income and the operating expenses are the key determinants of the transaction
values of building. We also introduce two indicator values for Energy Star ratings from the
earlier period of 2003 through 2005 and Energy Star ratings from the latest period, 2006
through 2009, when it is thought to be more difficult to get into the top twenty fifth percentile
of energy efficiency due to the increased competition for these scores among building owners.
As shown in Table 9 last two rows of the last column of the table, once appropriate controls
for net operating income, operating expenses, and capitalization rates are introduced into the
pricing equation, there appears to be no statistically significant additional effect of obtaining
an Energy Star label even in the most recent period.
These results suggest that the Energy Star rating can only be viewed as having a muted
effect on building transactions prices at least through the expense channel. Accounting for
the primary determinants of office building asset prices net rents, expenses, energy factor
prices and interest rates appears to leave no further room for either the “plaque-on-the-wall”
effect or the incremental savings associated with Energy Star certification. There is, of course,
one further control that is missing from these specifications and that is the cost associated
with actually obtaining an Energy Star rating. Given our result that the benefits of the
Energy Star certification appear to proxy for important missing factors in office building
asset prices, it unlikely that inclusion of these costs could change our conclusions that the
measurable net benefit of Energy Star is not statistically associated with transactions values
per square foot.
6 Leases
Important reductions in the energy consumption of U.S. buildings are technologically feasible,
but building owners-landlords-often do not receive the proper economic incentives to carry
out the required investments. The incentive failures occur in two related markets: the
rental market where lease contracts often inhibit energy efficiency, and the mortgage market
25
Table 9: Three Stage Least Squares Estimates for Endogenous Net Operating Income, Endogenous Total Operating Expenses,
and Prices for Office Buildings that Sold 2001 through 2010
This table presents the three stage least squares estimates for net operating income, total operating expenses per square foot, and prices
per square foot for office buildings that sold in arm-length-transactions between 2001 and 2010. We instrument net operating income
and operating expenses with exogenous energy factor prices, weather effects, time dummies and location dummies. The third stage least
squares estimates the price per square foot with instruments for endogenous net operating income and operating expenses.
Log Operating Log Net Operating Log Price
Expenses per Sq. Ft. Income per Sq. Ft. per Sq. Ft.
Parameter Standard Parameter Standard Parameter Standard
Variable Estimate Error Parameter Error Parameter Error
Intercept 1.217
∗∗∗
0.184 2.416
∗∗∗
0.193 4.938
∗∗∗
0.272
Six month lag price of the 1-12 month
forward contract at sale by transaction hub 0.001 0.002 0.002 0.002
Six month lag slope of the power forward curve -0.013 0.009 -0.007 0.009
Six month lag price of the 1-12 month
natural gas forward contract at sale -0.001 0.020 0.006 0.020
Six month lag slope of the natural gas forward curve .174
∗∗∗
0.089 0.131 0.093
Standard Deviation of 12 Month Maximum Temperature -0.002 0.002 -0.009
∗∗∗
0.003
Standard Deviation of 12 Month Minimum Temperature 0.035 0.003 0.008
∗
0.003
Standard Deviation of 12 Month Precipitation 1.434
∗∗
0.596 0.840 0.647
Number of floors 0.012
∗∗∗
0.001 -0.002 0.004 0.014
∗∗∗
0.005
Age of the Building -0.001 0.001 -0.001 0.001 -0.002
∗∗∗
0.001
Building square footage -0.003
∗
0.002 0.000 0.002 -0.002 0.002
Typical floor area square footage -0.015
∗∗
0.008 -0.018
∗∗
0.008 0.000 0.009
Indicator variable for class A buildings 0.055 0.044 0.053 0.046 0.173 0.050
Indicator Variable for Multi-Tenant Building 0.172
∗∗∗
0.042 0.047 0.044 0.157
∗∗∗
0.055
Indicator variable for renovation prior to sale 0.108 0.131
Market capitalization rate at year of sale -0.121
∗∗∗
0.010
Log Net operating income per square foot at year of sale 1.247
∗∗∗
0.131
Log Total expenses per square foot at year of sale -1.130
∗∗∗
0.125
Indicator variable for Energy Star rating
by time of sale (2003-2005) 0.285 0.178
Indicator variable for Energy Star rating
by time of sale (2006-2009) 0.030 0.289
Market Fixed Effects Yes Yes No
Year Fixed Effects Yes Yes Yes
R
2
Number of Observations 1470 1470 1470
26
where loan underwriting procedures also inhibit energy efficient investments. Lease contracts
contain many common basic terms and conditions that set the terms for payments and
services received between the tenant and landlord in three dimensions:
1. Space rent. The core purpose of a lease is to identify the space provided to the tenant
and the rent paid to the landlord. The lease will also typically identify the physical
condition of the space and any improvements the landlord will provide.
2. Building operating expenses. Building operating costs include energy use, property
taxes, building operations and maintenance, and insurance. Lease contracts will iden-
tify how the payments for these expenses are to be allocated between the landlord and
the tenants. Lease contracts will also typically indicate the quality level promised by
the landlord for building operations and maintenance.
3. Building capital expenditures. Building capital expenditures cover a variety of invest-
ments that maintain or improve the building, including investments to improve the
building’s energy efficiency. Lease contracts will identify how the amortized costs of
these investments are to be shared between the landlord and the tenants.
Lease contracts must also indicate the period over which the contract pertains. On longer-
term contracts, the lease will indicate how payments in the three categories will change over
time, quite possibly including how rising operating costs will be shared between the landlord
and the tenants. Furthermore, lease contracts may allow a variety of options such as allowing
either the landlord or tenant to break the lease under specified conditions.
Lease contracts for commercial buildings commonly take one of three main formats: full
service leases, net leases, and modified gross leases. Full service leases require that the
tenant makes a single payment that covers the tenant’s responsibility for space rent and
operating expenses. The individual components are typically not identified. This allows the
landlord freedom to reduce operating expenses, including energy costs, by making efficient
capital expenditures (subject to the minimum standard for building services and maintenance
specified in the lease contract).
With net leases, the tenant agrees to pay for both space rent and the tenant’s actual
or allocated share of the specified operating expenses. The operating expenses may include
energy, property taxes, and insurance-a lease that includes all three expense categories is
call a ”triple net” lease, but other combinations are possible. We focus here on the set of
Net leases that at least include electricity, which then requires that the tenant’s space have
direct metering.
Modified gross lease contracts specify a specific payment for the space rent plus and
stipulate an actual amount to be paid for operating expenses in the first year. For later
27
years, the landlord provides an audit of building expenses, and the tenants pays a prorated
share of the realized percentage increase in the building expenses. Modified gross leases
and Net leases share the feature that the tenant pays a share of the building’s operating
expenses, but on modified gross leases the tenant pays a prorated share of the building’s
total expenses, which are thus are independent of the tenant’s actual energy usage. For this
reason, modified gross leases are commonly used in buildings where energy metering of each
tenant’s space is not available.
6.1 Lease Contracts and Energy Efficient: Economic Theory
We begin with an outline of the economic theory of how lease contracts can affect energy
efficiency in office buildings. The assumed environment is that of a multi-tenant office
building. Lease contracts affect energy efficiency through two distinct channels: (1) tenant
and landlord energy use, taking as given the building’s current energy efficiency, and (2)
landlord investments that would improve the building’s energy efficiency. We focus first on
the energy use, taking as given the building’s current energy efficiency.
6.1.1 Contract Theory, Uncertainty, and Energy Use
Contract theory is the part of economics that studies the incentives received by contract
participants to take various actions. The specific implications depend critically on the en-
vironment in which the contracting process is assumed to occur. Uncertainty regarding a
building’s energy usage, including the possibility of asymmetric information between the
landlord and the tenants, is critical to understand the impact of lease contracts on energy
efficiency. In particular, if there were complete knowledge concerning the dollar costs of
energy usage, then lease contracts would have no impact on energy efficiency. This can
be illustrated by a simple example in which the space rent is $1,000 and the only relevant
building expense is the tenant’s energy use of $100. It is thus apparent that the tenant
will pay a total amount of $1,100. Under a Full Service or Modified Gross lease, the total
amount of $1,100 would be paid by the tenant to directly to the landlord. Under a Net
Lease, the tenant pays $1,000 to the landlord and $100 to the utility company, for the same
total payment of $1,00. It is thus apparent that the economic outcome is identical under the
three contracts.
The situation changes significantly once there is uncertainty concerning the likely energy
costs. For example, suppose there is a 50 percent chance the energy costs will be $50 and a
50 percent chance the energy costs will be $150, so the expected energy cost remains $100
as in the example. Under the Full Service contract, the landlord bears the risk that the
28
actual energy cost will be $150 and benefits if the actual cost is $50. Under a Net contract,
in contrast, the additional energy costs or benefits are realized by the tenant. The outcome
under a Modified Gross lease is potentially more complex, but assuming all tenants have the
same energy cost outcome, any additional energy costs or benefits will also be realized by
the tenant.
The implication is that lease contracts have an insurance component, with the risk of
high energy cost outcomes borne by the landlord under a Full Service lease and by the
tenant under a Net and Modified Gross leases. Economic theory indicates that the best
party to bear the risk-the tenant or the landlord-will be determined by which party is the
more risk tolerant. For example, if the landlord is the more risk tolerant, then we would
expect the market to adopt Full Service leases, thus allowing the tenants to avoid the risk
of unexpectedly high energy costs.
Asymmetric information-meaning that either the landlord or tenant has better informa-
tion concerning the likely energy costs-raises additional issues for the desired contracting
outcome. For example, if the landlord can estimate the likely energy costs with greater
precision than can the tenant, then this creates be a further reason for the landlord to bear
the risk as under a Full Service contract. On the other hand, when the tenant can control
the amount of energy use, and would significantly expand the use under a Full Service or
Modified Gross contract, then a Net contract may be the preferred contract outcome.
The implication is that Net lease contracts are commonly judged to be the most conducive
for energy efficiency because they provide tenants the greatest incentive to limit their energy
use. We will see in Section 3 below, however, that Net contracts are not a dominant outcome
for office building leases and that they are actually quite uncommon for large, multi-tenant,
office buildings in city centers. It is thus worth noting here several factors that limit the use
of Net leases:
1. Net lease contracts will not be effective where metering of individual spaces is not
available. Direct metering is limited because it is expensive and is even illegal in
certain jurisdictions.
2. Net lease contracts place all the risk of uncertain energy costs on the tenant. For situ-
ations in which the main uncertainty concerns the market price of energy, as opposed
to the physical amount used by the tenant, a Full Service contract may be preferred if
the landlord is more tolerant.
3. Net lease contracts may limit the incentive of a landlord to carry out energy efficient
building improvements, since the benefits of lower energy costs would then accrue, at
least initially, to the current tenants.
29
6.1.2 Capital Investments Under Alternative Lease Contracts
The preceding discussion has taken as given the building’s energy efficiency. Landlords, of
course, may carry out capital investments that would improve the building’s level of energy
efficiency. In this section, we consider the incentives landlords receive to carry out such
investments.
Net present value (NPV) is generally the preferred economic criterion for investment
decisions. Applied to an investment to improve a building’s energy efficiency, NPV would be
computed as the present value of the savings in future energy payments minus the investment
cost. The NPV criterion has the desirable feature that it provides the landlord an incentive
to carry out investments as long as the direct benefits exceed the direct costs. Of course,
there may also be societal benefits to carrying out energy efficient investments-for example,
to reduce global warming-and in these cases it may be useful for governments to provide
additional financial incentives. It is also worth nothing that rules of thumb such as Payback
Period-the number of years required to recover the initial investment costs-are often used as
approximations to a full NPV evaluation.
Whatever the detailed computational format, the standard application of NPV techniques
assumes that the landlord alone pays the costs and receives the benefits of the investment.
We have seen, however, that both the costs and benefits of energy investments may fall on
the tenants depending on the particular lease contract. The expected investment incentives
by contract form can be summarized as follows:
• Full Service Lease. Under Full Service leases, all the costs and benefits of energy
investments accrue to the landlord alone. Full Service leases are thus consistent with
the incentives provided landlords who apply an NPV criterion, and thus should provide
the optimal level of energy efficient investment levels.
• Net Lease. Under Net leases, the tenant both receives the benefits of reduced energy
costs and pays the amortized costs of the investment. Under ideal conditions, the
landlord and tenant will agree on the desired energy efficient investments, and thus the
same investments will be carried out as under a Full Service lease. One limiting factor,
however, is that the tenant’s occupancy horizon may not equal the expected economic
life of the capital investment. A second limiting factor is that the time patterns of the
amortized costs and the energy-saving benefits may differ. If, for example, the tenant
has a relatively short horizon and the amortization payments exceed the energy-saving
benefits over this period, then the tenant may object to the capital investment even
when it is NPV positive. A further issue also arises in the common case when the
tenant move outs before the capital investment is fully amortized. In this case, the
30
landlord will have to convince the new tenant to pay a higher space rent based on lower
expected energy costs. If the landlord fears she will be unable to achieve a higher rent,
then she may be reluctant to carry out the investment in the first place. On the other
hand, assuming the investment is NPV positive and that the landlords of comparable
buildings carry out such investments, then a landlord who fails to make the investment
runs the risk of lowing potential tenants and receiving only lower rents.
• Modified Gross Lease. Modified Gross leases are comparable to Net leases in that
tenants receive the benefits of reduced energy costs, and thus, at least to a first approx-
imation, landlords face the same incentives with regard to energy efficient investments.
Table 10 presents the contractual structures that are represented in the CoStar data. As
shown, we are missing information on the dominant leasing structure for more than 57%
of the transactions. For those buildings for which we have leasing structure, the dominant
types are the full service gross with 18.8% of transactions, where the landlord pays all the
energy costs and the tenants pay a pro rata share based on their percentage of leased space,
and some form of triple net lease, 18.04% where the tenant pays the costs.
Table 10: Lease Contract Types in the COSTAR Office Building Transaction Data
This table presents the primary lease contracting methods for allocating the utility expenses (water,
natural gas, and electricity expenses) to tenants in the COSTAR building that traded in arms-
length-transactions between 1999 and 2010. These data were obtained by merging the COSTAR
rental data with the COSTAR Transactions data.
Contract Type Frequency Percent Cumulative Frequency Cumulative Percent
Full Service Gross 2768 18.29% 2768 18.29%
Modified Gross 874 5.78% 3642 24.07%
Triple Net 2626 17.35% 6268 41.42%
Missing 8865 58.58% 15133 100.00%
There is also important geographic structure to the location of office buildings that rely on
different leasing structures. For example, as shown in the Appendix in Figure 3 and Figure 4,
the Net Leases contracts predominate in the South Bay (Silicon Valley) area where office
buildings often include an important component of tenants characterized by high electricity
consumption required for “clean rooms” and computers. Net lease are also found in other
peripheral office market areas in both the San Francisco and Los Angeles markets. The full
service and modified gross leases are found primarily in central business district areas in the
31
Bay Area and Los Angeles office markets.
6.2 Empirical Tests
At least one of the implications of the preceding discussion is that energy use in Triple Net
buildings would be lower reflecting the direct incentive of tenants to minimize their energy
bill. Although we cannot directly test this hypothesis, we consider two specifications to
explore these effects: an estimation of the forward looking asset pricing specification defined
by Equation (2) with controls introduced for the lease contracting structure of the building; a
re-estimation of the total operating expenses relationship again with controls for the leasing
structure.
As shown in Table 11, the results for the valuation expression are very similar to those
reported for specifications for Equation (2), reported in Table 7, that did not include controls
for the lease contracting structure of the building. Given the small sample size, the slope
of the power forward curve has a statistically significant and positive effect on the price
per square foot but only at the 10% level of statistical significance. The other energy factor
prices no longer have statistically significant effects on prices. The weather effects retain their
strong and statistically significant effects on log price per square foot even after controls for
fixed effects, although their effects of variance in minimum and maximum temperatures
appear to countervail each other. Again, the importance of the Equation (2) specification
for asset prices is borne out with the results for net operating income, the capitalization rate,
total expenses per square foot are all economically and statistically significantly associated
with log transaction prices per square foot. Again, the indicator variable for an Energy Star
certification at time of sale contributes no additional explanatory power in this regression,
nor does a specification that accounts for the period in which the Energy Star certification
was received.
The introduction of the indicator variables for predominance of Full Service and Modified
Full Service Leases in the building is shown to have a statistically and significantly negative
effect on transaction prices relative to the omitted category of Triple Net Leases. Indeed,
this result appears to suggest, as expected from the incentive structure of Full Service and
Modified Full Service Leases that contractual inducements for tenants to minimize the utility
costs of their space-use does appear to affect building value. Since utility costs are paid as
a pro rata share in Full Service and Modified Full Service Leases, individual tenants have
little incentive to minimize their utility consumption. This differential incentive structure
appears to have important effects on transaction values at the margin.
In Table 12, we report the results of re-estimating the log of total operating expenses
32
Table 11: Regression Results for Office Buildings that Sold 2001 through 2010
This table presents the regression results for the log price per square foot at sale buildings that
sold in arm-length-transactions between 2001 and 2010. The dependent variable is the log price
per square foot and it is regressed on energy price and weather effects, building characteristics,
leasing characteristics, and the Energy Star rating status of the building at the time of sale. These
data were obtained from the COSTAR Transactions data base. The energy forward curve data was
estimated using data from Platts and NYMEX and the weather data was obtained from Wolfram
Schlenker.
Parameter Standard Parameter Standard
Variable Estimate Error Estimate Error
Intercept 5.418 0.248 5.416
∗∗∗
0.249
Six month lag price of the 1-12 month
forward contract at sale by transaction hub 0.004 0.003 0.004 0.003
Six month lag slope of the power forward curve 0.016
∗
0.008 0.015
∗
0.008
Six month lag price of the 1-12 month
natural gas forward contract at sale -0.007 0.013 -0.005 0.013
Six month lag slope of the natural gas forward curve -0.100 0.080 -0.095 0.080
Standard Deviation of 12 Month Maximum Temperature -0.011
∗∗∗
0.003 -0.011
∗∗∗
0.003
Standard Deviation of 12 Month Minimum Temperature 0.011
∗∗∗
0.003 0.010
∗∗∗
0.003
Standard Deviation of 12 Month Precipitation 0.036 0.633 0.039 0.632
Number of floors 0.000 0.003 0.001 0.003
Age of the Building -0.004
∗∗∗
0.001 -0.004
∗∗∗
0.001
Building square footage 0.000 0.001 0.000 0.001
Typical floor area square footage 0.007 0.007 0.007 0.007
Indicator variable for renovation prior to sale -0.014 0.138 -0.012 0.138
Indicator variable for class A buildings 0.203
∗∗∗
0.034 0.199 0.034
Indicator Variable for Multi-Tenant Building 0.013 0.054 0.013 0.054
Market capitalization rate at year of sale -0.098
∗∗∗
0.011 -0.098
∗∗∗
0.011
Total expenses per square foot at year of sale -0.199
∗∗∗
0.047 -0.199
∗∗∗
0.047
Net operating income per square foot at year of sale 0.501
∗∗∗
0.045 0.502
∗∗∗
0.045
Predominate Lease Structure: Full Service Gross -0.154
∗∗∗
0.039 -0.154
∗∗∗
0.039
Predominant Lease Structure: Modified Gross -0.213
∗∗∗
0.056 -0.214
∗∗∗
0.056
Indicator variable for Energy Star rating by time of sale 0.077 0.079
Indicator variable for Energy Star rating by time of sale (2003-2005) -0.004 0.001
Indicator variable for Energy Star rating by time of sale (2006-2009) -0.009 0.292
Market Fixed Effects Yes Yes
Year Fixed Effects Yes Yes
Clustered Standard Errors (Sale date and Market) Yes Yes
R
2
0.668 0.667
Number of Observations 558 558
33
Table 12: Regression Results for Total Operating Expenses for Office Buildings that Sold
2001 through 2010
This table presents the regression results for the total operating expenses of office buildings that
sold in arm-length-transactions between 2001 and 2010. The dependent variable is the log operating
expense per square foot at sale and it is regressed on energy price and weather effects, building
characteristics, leasing characteristics, and the Energy Star rating status of the building at the time
of sale. These data were obtained from the COSTAR Transactions data base. The energy forward
curve data was estimated using data from Platts and NYMEX and the weather data was obtained
from Wolfram Schlenker.
Parameter Standard
Variable Estimate Error
Intercept 1.304
888
0.188
Six month lag price of the 1-12 month
forward contract at sale by transaction hub -0.001 0.002
Six month lag slope of the power forward curve 0.004 0.007
Six month lag price of the 1-12 month
natural gas forward contract at sale -0.015 0.012
Six month lag slope of the natural gas forward curve -0.015 0.054
Standard Deviation of 12 Month Maximum Temperature -0.005 0.003
Standard Deviation of 12 Month Minimum Temperature 0.008
∗∗
0.003
Standard Deviation of 12 Month Precipitation 1.150
∗∗
0.562
Number of floors 0.012
∗∗∗
0.003
Age of the Building 0.000 0.001
Building square footage -0.002 0.001
Typical floor area square footage -0.006 0.006
Indicator variable for renovation prior to sale -0.152
∗
0.091
Indicator variable for class A buildings 0.145 0.034
Indicator Variable for Multi-Tenant Building 0.257
∗∗∗
0.045
Predominate Lease Structure: Full Service Gross 0.015 0.040
Predominant Lease Structure: Modified Gross -0.050 0.056
Indicator variable for Energy Star rating by time of sale -0.077 0.070
Market Fixed Effects Yes
Year Fixed Effects Yes
Clustered Standard Errors (Sale date and Market) Yes
R
2
0.321
Number of Observations 953
34
on market and building characteristics when we introduce controls for the dominant lease
contracting structure of the building. As shown, overall, the results are similar to those
discussed in the preceding sections and we find that multi-tenant building are economically
and statistically associated with higher total operating expenses. Interestingly, the controls
for the Full Service and Modified Full Service Leases are not associated with the level of
total operating expenses per square foot nor is the Energy Star Certification. Given the
small sample size, it is difficult to draw strong conclusions about the relationship between
building costs and the leasing structure of buildings. Nor it is possible, to clearly differentiate
the cost effects of leasing structure of the office buildings in our sample.
7 Conclusions
This paper presents an empirical analysis of the relationship between energy factor markets,
leasing structures and the transaction prices of office buildings in the U.S. We employ a
large sample of 15,133 office building transactions that occurred between 2001 and 2010. In
addition to building characteristics, we also include information on the operating expenses,
the net operating income, and the capitalization rates at sale to estimate an asset pricing
model for commercial office real estate assets. A further set of important controls in our
analysis is the one-to-twelve month forward contract prices and the shape of the forward
contract price curve, using auction data from the major electricity trading hubs in the
U.S. and from the Henry Hub for natural gas. We also include weather metrics in the
form of the variance in the last twelve months of minimum and maximum temperature and
precipitation from each building’s sale date. Our final set of controls includes information on
the dominant leasing structures in the buildings. Our empirical results suggest that Energy
Star labels do not explain additional variance in property prices once the key asset pricing
factors of expenses, income and capitalization rates are included. Energy factor market
prices, the shape of the energy price curves, and weather metrics are consistently shown to
be statistically significant determinants of office building transaction prices, suggesting that
commercial office building prices are likely to be exposed to shocks in these markets. This
finding has important implications for underwriting commercial mortgage default risk.
35
A Appendix
A.1 Geographic Structure of the Transactions Data
A.2 Weather Data Construction
The weather data were obtained from Wolfram Schlenker at Columbia University. These data
are based on the same rectangular grid system underling PRISM that covers the contiguous
United States
18
It consists of 1405 grids in the longitude direction and 621 grids in the
latitude dimension, space equidistant 1/24 degree steps (about 2.5 miles). The data are
matched to the centroid of each grid point to the fips codes of all counties in the United
States. There are 471,159 grid points with non-missing data in the PRISM data where the
centroid is matched to lie within a county. These are displayed in shades of grey in Figure 7.
There were 390,864 grid points that had missing values in PRISM and the centroid could not
be matched to a fips code. These grids are display in blue in Figure 7. There are 10,480 grid
points with non-missing valued in PRISM whose centroid could not be matched to a county.
These are primarily around the border of the United States as well as over water bodies like
the Chesapeake or San Francisco Bay. They are displayed in red in Figure 7. There were
also 2 grids that had missing values in PRISM but whose entroids could be matched to a
county. They are displayed in green in Figure 7 and can be found at the shoreline of the
Great Lakes.
The data include the minimum and maximum temperature (Celsius), total precipitation
(cm) for each day of a year for all of the 471159 grids in the United States from 1950 through
2010. These data are interpolated from PRISMs monthly weather station averages to daily
data and we aggregate them back into monthly data for our analysis. We associate the past
twelve months of weather data for each building in the COSTAR data with the weather data
associated with the nearest grip point in the Schlenker, et al. data.
A.3 Energy Data Construction
We extract the energy forward curve pricing from the forward contract auctions for electricity
and from the futures contracts auctions for natural gas. We follow Benth, Koekebakker,
and Ollmar (2007), Benth, Cartea, and Kiesel (2008), Geman and Roncoroni (2006) and
Riedhauser (2000), in the construction of these curves.
18
http://www.prism.oregonstate.edu/
36
A.3.1 Forward Market for Power (Electricity)
The forward market for power is organized around the trading of different standard packages
covering the on-peak and off-peak periods. Trading occurs for delivery hubs located at the
Eastern-Central regions and delivery hubs located in the Western region of the continental
United States. The Easter-Central standard forward package covers the following markets:
New England, New York (several hubs), Ontario, PJM, MISO, ERCOT South, Into Entergy,
Into Southern and Into TVA. The Western packages cover NP15 and SP15 among others.
Packages for the Eastern-Central hubs differ from those traded for the Western hub on two
dimensions: the way on-peak and off-peak are defined and the delivery months of the forward
packages.
We compute the standard on-peak forward packages in Eastern and Central markets are
5x16 packages (5 days per week and 16 hours per weekday from 7:00 Am to 22:59 PM),
which include power delivered during on-peak hours on weekdays and exclude weekends and
holidays .
19
Similarly, on-peak forward packages in Western markets are 6x16 packages, which
include power delivered during the 16 on-peak hours each day Monday through Saturday
and exclude Sundays and holidays. The off-peak standard packages, the forward market
trade 5x8 (5 days per week and 8 hours per day) plus a 2x24 package, this includes power for
delivery during the eight off-peak hours each weekday, plus all 24 hours (around the clock) on
weekends. The standard off-peak forward package for the Western markets is a 6x8 delivery
block plus a 1x24 delivery block, this includes power for delivery during the eight off-peak
hours Monday through Saturday plus all 24 hours (around the clock) on Sunday.
For the Eastern-Central markets, on-peak and off-peak contracts are formulated for the
prompt month (nearest contract), second month, third month, and balance-of-the-year in
seasonal or single month packages, two full years in seasonal or single-month packages and
two subsequent calendar year packages. Separate seasonal and single-month packages include
the January-February winter package, the March-April spring package, May, June, the July-
August summer package, September and the fourth quarter (from October to December).
The following example illustrates, for a given trading date, the typical term-structure
of contracts for packages traded in the Eastern-Central hubs. Suppose the today’s date is
5/15/2009. At this time, the market is trading the following forward packages:
• 2009 - June-2009 (prompt month), July-2009 (second month), August-2009 (third
month), September-2009 (single month package), fourth-quarter 2009,
• 2010 - January-February-2010 (winter package), March-April-2010 (spring package),
May-2010 (single month package), June-2010 (single month package), July-August-
19
Power market holidays are defined by the North American Electric Reliability Corp. (NERC).
37
2010 (summer package), September-2010 (single month package), fourth-quarter 2010,
• 2011 - January-February-2011 (winter package), March-April-2011 (spring package),
• May-2011 (single month package), June-2011 (single month package), July-August-
2011 (summer package), September-2011 (single month package), fourth-quarter 2011,
• 2012 - year-2012 (calendar year package), 2013 - year-2013 (calendar year package).
For the Western markets, on-peak and off-peak packages are formulated for the prompt
month, second month, balance of the year in quarters, two full years in quarters, and two
subsequent calendar year packages.
As before, suppose the today’s date is 5/15/2009. At this time, the market is trading
the following forward packages for a Western hub:
• 2009 - June-2009 (prompt month), July-2009 (second month), August-2009 (third
month), third-quarter 2009 (July-September package), fourth-quarter 2009 (October-
December package),
• 2010 - first-quarter 2010 (January-March package), second-quarter 2010 (April-June
package), third-quarter 2010 (July-September package), fourth-quarter 2010 (October-
December package),
• 2011 - first-quarter 2011 (January-March package), second-quarter 2011 (April-June
package), third-quarter 2011 (July-September package), fourth-quarter 2011 (October-
December package),
• 2012 - year-2012 (calendar year package),
• 2013 - year-2013 (calendar year package),
A significant portion of transactions are realized over-the-counter (OTC). Transactions
also occur on specialized exchanges such as the IntercontinentalExchange (ICE). Often trad-
ing parties take their existing OTC transactions for clearing into ICE. This mechanism
mitigates counterparty risk since parties are now at arms-length and are subject to margin
calls as prices for the forward packages fluctuate.
The energy volume for a typical package is 50 MWh (Megawatt-hour) times the number
of on-peak or off-peak hours depending on the type of package. If for example, the parties are
trading an on-peak September-2011 contract delivering into the PJM Western hub. Assuming
that there are 22 NERC weekdays for this month. The total volume for such a package would
is then calculated as (50 x 22 x 16) MWh.
Parties engage in financial settlement a few days after the ending date of the package. As
in the example above, the settlement for the September-2011 package occurs in the beginning
of October-2011. The long party receives (pays) the difference between the floating price
(calculated, in this case, as the arithmetic average of the arithmetic averages of the hourly
38
on-peak real-time prices posted by PJM Interconnection, LLC, on their official website) and
the agreed forward price at the time of contracting.
20
A.3.2 Dataset - Platts-Ice Forward Curve
In this section we describe the source of data for constructing our power forward price level
and slope database. We contracted the raw data for Platts (the data vendor) and worked
out a daily forward curve for power on on-peak and off-peak comprising selected trading
hubs serving large cities in The United States.
Platts gathers information on the power forward market from active brokers and traders
and through the non-commercial departments (back offices) of companies. Since October
2007 this information is complemented with the IntercontinentalExchange (ICE) quotes to
form the Platts forward market power daily assessment. Since more liquid locations and
shorter term packages trade more on ICE, while less liquid locations and longer term pack-
ages trade more over-the-counter (OTC), Platts is able to combine these sources to build
a comprehensive picture of the forward market. Details of the methodology are described
in the Platts Methodology and Specification Guide - Platts-ICE electricity Forward Curve
(North America).
In our analysis, we selected a subset of trading hubs corresponding to large urban pop-
ulated areas covering both the eastern, western, southern regions of the United States. The
following table describes these power hubs, the related urban areas referenced by the hub,
and the closest wholesale natural gas trading hub.
Figure 6 presents the geographic location of all the the power hubs in the U.S. We select
a sub-set of hubs based on data availability for options and forward contracts from Platts
and by our requirement to account for the power forward prices for all metropolitan areas
with 150,000 employees in Finance, and Professional and Business Services (the major office
categories).
21
As shown, in the Figure there is considerable regional variation in the level
of power forward prices and in their time series dynamics. The hub locations correspond to
the nodal structure of the natural gas pipeline in the U.S. and to the location of the major
population centers in the U.S.
Raw data from Platts is formatted with single entries for each forward package (see the
”Forwards Market for Power” section above). For a given trading date, a power hub, and
a type of contract - on peak and off-peak - there are single entries for the mark-to-market
price for each forward package. This scheme characterizes a whole term-structure of power
20
We refer the reader to the Intercontinental Exchange website for more details on how the contracts are
traded and settled.
21
Bureau of Labor Statistics, Employment Hours and Earnings, State and Metro Area, http://www.bls.
gov/sae/data.htm
39
prices for a given trading date. The following table describes for each hub and contract type
the time series of the related forward curves.
Table 13: The Earliest and Latest Trading Dates for the Power Hubs
This table presents the periods for which we have data on the forward and spot prices for each
electricity hub.
Contract Type Region Name Minimum Trade Date Maximum Trade Date
on-peak East New York Zone J 1/2/2002 4/23/2010
on-peak ERCOT 1/2/2002 4/27/2010
on-peak Mass Hub 8/30/2002 4/23/2010
on-peak NI Hub 1/2/2002 4/23/2010
on-peak North Path 15 12/31/2001 4/23/2010
on-peak PJM West 3/5/2001 4/23/2010
on-peak South Path 15 12/31/2001 4/23/2010
on-peak Into Cinergy 3/1/2001 10/23/2010
on-peak Into Entergy 3/1/2001 10/23/2010
on-peak Into Southern 12/1/2005 10/23/2010
on-peak Into TVA 1/1/2002 10/23/2010
on-peak Mid Columbia 3/1/2001 10/23/2010
on-peak Palo Verde 3/1/2001 10/23/2010
off-peak East New York Zone J 1/31/2007 4/27/2010
off-peak Mass Hub 2/7/2007 4/23/2010
off-peak NI Hub 1/31/2007 4/23/2010
off-peak North Path 15 5/31/2006 4/23/2010
off-peak PJM West 1/31/2007 4/23/2010
off-peak South Path 15 5/31/2006 4/23/2010
off-peak Into Cinergy 3/1/2001 10/23/2010
off-peak Into Entergy 3/1/2001 10/23/2010
off-peak Into Southern 12/1/2005 10/23/2010
off-peak Into TVA 1/1/2002 10/23/2010
off-peak Mid Columbia 3/1/2001 10/23/2010
off-peak Palo Verde 3/1/2001 10/23/2010
After inspecting the raw data, we noted the following:
1. We did not find any significant gaps on trade dates for all time series.
2. The time series for on-peak New York Zone-J has gaps on the term-structure from the
beginning of the time series 1/2/2002 to 1/11/2005. Consequently, we discarded these
raw entries when calibrating instantaneous volatilities.
40
3. The ERCOT (all zones) time series has data up to 11/25/2008. Consequently, we
extended the time series for ERCOT by appending the ERCOT-South time series
starting in 11/26/2008.
4. We discarded raw entries with trade dates later than the beginning of the delivery
period. For example, we found on 3/6/2008 a quote for the on-peak East NY ZnJ
Mar/Apr 2008 package. Note that on this trade is already into the delivery period
of the contract which starts on 1/3/2008. Though these types of trades are valid,
the quote corresponds to parties trading on information related to the balance of the
delivery package. Consequently, since the structure of the contract is now different
from the original package, we discard these entries.
5. The length of the forward curve increases for more recent years. We illustrate this by
showing below the maximum number of months out by trading year for the on-peak
PJM Western hub and NP15 hub.
The raw data from Platts contains a field called ”symbol” indicating the type of contract,
the package, the power hub, and the year related to the package. The ”symbol” field is coded
with 7 characters with the following structure and the last two characters is a two digit code
for the hub name. The other codes include:
The following describes Platts coding for the sub-fields seasonal package and the Contract
Length:
A.3.3 Futures Market for Natural Gas
There is a very active market for natural gas in The United States. Following deregulation
of the wholesale market for natural gas in mid 1990’s, the New York Mercantile Exchange
(NYMEX) launched the trading of monthly futures contracts with similar characteristics
to those of crude oil. The standard NYMEX natural gas futures contracts specify physical
delivery of 10,000 MMBtu (millions of British thermal unit) ratably delivered into Henry
Hub - Louisiana. Until early 2000’ NYMEX provided monthly contracts covering maturities
about 36 months out. More recently, the range of maturities has been extended and it now
covers more then six years (72 months) on a monthly basis. The NYMEX website provides
more details on how the contracts are traded and the rules for settlement.
The following graph shows NYMEX’s prompt month (nearest contract) daily quotes
starting in 2005.
There is an extensive network of natural gas pipelines connecting the production basins to
large consumption areas (mainly large populated urban centers). Wholesale physical natural
gas trading occurs in different hubs distributed in the continental United States. These hubs
41
Table 14: The Maximum Number of Months out by Trading Year for the On-peak PJM
Western and NP15 Hubs
This table shows that the length of the forward curve has increased for more recent years.
Region Name Year of Trade Date maximum Month Out
North Path 15 2001 35
North Path 15 2002 41
North Path 15 2003 41
North Path 15 2004 41
North Path 15 2005 59
North Path 15 2006 59
North Path 15 2007 59
North Path 15 2008 59
North Path 15 2009 59
North Path 15 2010 59
PJM West 2001 41
PJM West 2002 41
PJM West 2003 41
PJM West 2004 41
PJM West 2005 41
PJM West 2006 47
PJM West 2007 59
PJM West 2008 59
PJM West 2009 59
PJM West 2010 59
Table 15: The Symbol Codes from Platts
This table presents the symbol code keys from Platts that are used to identify the class of forward
and spot contracts and the location of the trading hub.
Position sub-field size Sub-field indicator comments
1-Jan 1 on-peak / off-peak “F” for on-peak and “O” for off-peak
3-Feb 2 seasonal package see table below
5-Apr 2 power hub name see table below
7-Jun 2 package year —
42
Table 16: Codes for the Sub-Field Seasonal Packages and their Contract Length
This table presents the codes used by Platts to identify seasonal packages, their start and end dates,
and the forward contract length.
Contract Code Contract Description Start Month End Month contract Length
(seasonal package)
AA January 1 1 1
AB February 2 2 1
AC March 3 3 1
AD April 4 4 1
AE May 5 5 1
AF June 6 6 1
AG July 7 7 1
AH August 8 8 1
AI September 9 9 1
AJ October 10 10 1
AK November 11 11 1
AL December 12 12 1
AN January-February 1 2 2
AP March-April 3 4 2
AT July-August 7 8 2
AY Year 1 12 12
Q1 First Quarter 1 3 3
Q2 Second Quarter 4 6 3
Q3 Third Quarter 7 9 3
Q4 Fourth Quarter 10 12 3
43
are key points in the pipeline grid characterized by either being interconnections between
major pipelines and/or access points to public utility gas companies. Of all those hubs,
Henry Hub is the benchmark for price quotation. Henry Hub’s importance stems from both
as being an interconnecting point for multiple pipelines and as being the most liquid point
for trading spot and futures contracts. Prices for other hubs (spot and OTC forwards) are
typically quoted as a basis to Henry Hub. These basis quotes are a very small fraction of
the full benchmark quote.
44
Figure 1: EPA Energy Star Rated Buildings Located in the Los Angeles, Riverside, and San
Diego Areas
Red Dots: Non EnergyStar Rated Buildings Blue Dots: EnergyStar Rated Buildings
Figure 2: EPA Energy Star Rated Buildings Located in the San Francisco, East Bay, San
Jose, and Sacramento Market Area
Red Dots: Non EnergyStar Rated Buildings Blue Dots: EnergyStar Rated Buildings
45
Figure 3: Lease Contract Types for Building Locations in San Francisco, East Bay, San Jose,
and Sacramento Area
Green Dots: Full Service Leases Blue Dots: Modified Gross Leases
Red Dots: Triple Net Leases
46
Figure 4: Lease Contract Types for Building Locations in Los Angeles, Riverside, and San
Diego Area
Green Dots: Full Service Leases Blue Dots: Modified Gross Leases
Red Dots: Triple Net Leases
47
Figure 5: Geographic Grids in the Schlenkman et. al. Weather Data
48
Figure 6: Federal Energy Regulatory Commission Geographic Location of the Power Hubs
in the United States
This figure was obtained from the Federal Energy Regulatory Commission
(www.ferc.gov/oversight). It presents the geographic location of the hubs for
electricity forward contract auctions in the U.S.. The average dollar value of
the near contract over the year 2009 is presented for each hub and the pre-
centage change in this average price from the average over the year 2008.
Figure 7: NYMEX Natural Gas Prompt Month (nearest contract) Daily Quotes
This figure was computed from daily quotes for NYMEX nat-
ural gas prompt month (nearest contract futures contracts.
49
References
Benth, F., A. Cartea, and R. Kiesel, 2008, Pricing forward contracts in power markets by
the certainty equivalence principle: Explaining the sign of the market risk price premium,
Journal of Banking and Finance 32, 2006–2021.
Benth, F. E., S. Koekebakker, and F. Ollmar, 2007, Extracting and applying smooth for-
ward curves from average-based commodity contracts with seasonal variation, Journal of
Derivatives 15, 52–66.
Geman, H., and A. Roncoroni, 2006, Understanding the fine structure of electricity prices,
Journal of Business 79, 1225–1261.
Grenadier, S. R., 2005, An equilibrium analysis of real estate leases, Journal of Business 78,
1173–1214.
Plazzi, A., W. Torous, and R. Valkanov, 2010, Expected returns and hte expected growth in
rents of commercial real estate, Review of Financial Studies 23, 3469–3519.
Riedhauser, C., 2000, A no-arbitrage forward commodity curve, Working paper, Pacific Gas
and Electric Company.
Wheaton, W. C., and R. Torto, 1994, Office rent indices and their behavior over time,
Journal of Urban Economics 35, 121–139.
50
