Producer and consumer responses to green housing labels.
Shewmake, Sharon ; Viscusi, W. Kip
I. INTRODUCTION
Quality ratings are of particular value in situations in which
consumers cannot monitor and producers cannot credibly communicate the
quality of the product. Environmental rating systems for houses focus
primarily on energy efficiency but often include broader environmental
externalities. We refer to such labeling systems as green labels. These
labels serve as quality rating systems that provide energy cost
information to consumers and serve a broader policy objective of
fostering energy conservation, reducing water consumption, minimizing
construction waste, and providing a healthy indoor air environment.
There are many ways such labels might convey quality (metrics such as
miles per gallon, seals of approval such as dolphin safe, or gradations
of quality such as 1, 2, or 3 stars).
This paper finds that the structure of the rating system can
influence both the types of homes available and the market premiums for
homes with different gradations of green ratings. The findings in this
article suggest that labeling authorities may lead consumers to pay more
for higher quality, in this case greener, products by adopting a
labeling policy that provides information at different levels of green
stringency. Our conclusion is based on estimates using information from
three green labeling programs for houses in Austin, Texas. We assess the
differential role of such alternative labeling policies across the
housing market and find the greatest effects of stringent green ratings
at the high end of the market. Notably, the supply of green housing
attributes is clustered at or just above the critical green gradation
cutoffs. This targeting by builders is a reasonable response to the
labeling policies as there is no statistically significant price effect
of green "points" within particular green label categories.
For a given stock of homes, the effect of energy labels on house
prices can be captured by applying the insight of "lemons"
models in which there is the opportunity for quality certification
(Akerlof 1970; Blackman and Rivera 2010; Brouhle and Khanna 2007; Hotz
and Xiao 2013; Viscusi 1978). Energy labels serve as a quality rating
system with respect to energy efficiency or greenness of the house.
There are many energy efficiency aspects of houses that are difficult
for consumers to monitor. The insulation behind a wall is not visible,
and even if the consumer can observe that a house has low emissivity
windows, the consumer may not know the extent to which these windows
affect cooling bills. For a mix of houses that is relatively homogeneous
and is viewed as being of comparable quality other than the unobservable
differences in energy efficiency, the incentives for conveying the
quality information will be greatest for the houses that offer the
greatest energy efficiency savings relative to a comparison set of
houses. Thus, we would often expect the most energy efficient houses to
have the greatest incentives to provide the quality information and less
energy efficient homes to have diminished incentives to communicate
their green rating. Whether the incentives to certify the quality level
always decline for houses further down the quality spectrum also may
depend on the distribution of quality in the market, as this will affect
the perceived difference in average quality conditional on a quality
rating.
Unlike the standard lemons models in which there is a fixed stock
of cars of uncertain quality, in the case of energy efficiency ratings
for homes, builders can invest in energy efficiency or other green
components to alter the supply of homes at different quality ratings. If
there is a positive income elasticity of demand for green home
attributes, the demand for green ratings may be higher for more
expensive homes. The cost structure for achieving a green rating also
may vary across the market, but the direction of the effect is unclear.
Larger homes may entail more costs to achieve the green rating, but more
expensive homes may already include additional amenities reducing the
cost of achieving a green quality rating. Because of the influence of
such demand and supply effects, the distribution of green label
certifications may not be uniform across the market. Similarly, if there
are gradations of energy efficiency certifications for which the costs
of achieving the certification rise with the efficiency level, there
likewise may be an uneven distribution of these high efficiency ratings.
Communicating the energy usage of a home is difficult even when one
has knowledge of how a building's architecture influences energy
efficiency. Recent work on how consumers make data-intensive decisions
suggests an evaluative metric (e.g., poor, average, good on a 1-5 star
scale) reduces cognitive hurdles and helps consumers pick better higher
quality products at similar prices (Hibbard and Peters 2003; Hibbard et
al. 2012). Although consumers may find it easier to make decisions when
scales have well designed evaluative categories, many qualitative scales
are confusing or poorly understood.
The housing data we examine in this paper are particularly rich in
that there are several green label programs that we can analyze.
Moreover, one of these programs provides rating for different gradations
of energy efficiency. After describing the data and the different green
labeling programs in Sections II and III, Section IV documents the
clustering of the ratings just above the cutoffs for higher green
ratings in a manner that is consistent with evidence of targeting a
ratings notch in other contexts, such as miles per gallon (mpg) ratings
for cars (Sallee and Slemrod 2012). We present hedonic price regression
results for the market overall in Section V. Houses with green labels
command a price premium, but homes that obtain points beyond label
thresholds do not receive any further price premiums. We use quantile
regression in Section VI to explore the effects across the market. The
average effect of having some kind of green label increases the value of
a home by approximately 5%. More stringent labels generate greater price
effects. One of the labeling systems provides information on gradations
of environmental performance so that consumers will be able to
distinguish levels of green houses, with the houses in the highest
category receiving a 20% premium. Analysis of the differential role of
green ratings across the market indicates that the distribution of these
ratings across the market is not uniform, as the upper end homes are
more likely to achieve the most stringent ratings. The large percentage
price effect of a high green rating implies a large absolute price
effect for these homes, which also makes the investment in energy
efficiency more financially attractive.
II. ENERGY RATINGS FOR HOMES
A. Background on Energy Rating Studies
Currently, there are multiple green certifications available for
homes in the United States. Some of these certifications focus on one
dimension such as energy efficiency while others incorporate multiple
environmental attributes. In Austin, Texas, at least seven competing
green certifications are available to homebuyers. These programs vary
with respect to the attributes they consider--energy efficiency only
versus a multidimensional version of "green," how they
communicate the certification--a tiered approach with varying levels of
green versus a single threshold that comes with a "seal of
approval," and the party that is doing the certification, such as
the federal government, local government, nonprofits, industry
associations, or a private corporation. This article focuses on the
three programs that are most common in Austin: the Austin Energy Green
Building (AEGB) program, the Energy Star Homes program, and the
Environments for Living (EFL) program. These certifications are not
mutually exclusive. Many homes have multiple ratings, but each rating
requires additional testing, paperwork, and fees to the rating agency.
(1) We focus on results for the AEGB homes, but compare these results to
the Energy Star Homes and EFL programs.
Previous work has found that green commercial buildings typically
receive a price premium of between 6% and 67% depending on the type and
level of certification (Chegut, Eichholtz, and Kok 2012; Eichholtz, Kok,
and Quigley 2010, 2013; Fuerst and McAllister 2011; Harrison and Seiler
2011). There is also evidence of positive price premiums for green
certifications of residential buildings in California (Kok and Kahn
Forthcoming) and the Netherlands (Brounen and Kok 2011). A related
literature measures consumers' willingness to pay for energy
efficient houses. While this literature does not consistently define
energy efficiency, the price premiums consumers are willing to pay for
homes that use less energy are positive (Aroul and Hansz 2011; Johnson
and Kaserman 1983; Lande 2006; Nevin and Watson 1998). Previous work on
the impact of the U.S. Green Building Council's (USGBC) Leadership
in Energy and Environmental Design (LEED) certification on commercial
buildings has found that more stringent certifications result in higher
premiums (Fuerst and McAllister 2011).
Our paper extends these approaches, considering the effect of
multiple green labels and gradations in green labeling on the U.S.
private home prices as well as the targeting housing supply effects of
ratings categories. We find the structure of the labeling system matters
in terms of creating discontinuous incentives for producers and having
differential effects across the housing market. Additionally, we control
for important green building incentives such as an area that can only be
developed with green certifications and the interaction of low-income
housing. Similar to findings for the commercial real-estate market, we
find that the premium for green residential homes increases with green
certification stringency.
While we examine the market impacts of various green labeling
policies, we do not evaluate whether the impacts enhance economic
efficiency. When asked to evaluate the air pollution impacts of cars
with a green seal of approval instead of more quantitative information
on emissions, consumers consistently overestimated the gains from the
green seal of approval (Teisl, Rubin, and Noblet 2008). The supply side
response to evaluative categories may produce additional distortions. In
their study on vehicle mileage efficiency, Sallee and Slemrod (2012)
found that producers had large incentives to increase mpg when they were
just below a threshold, and this led to large losses in efficiency.
Thus, the presence and usage of an energy efficiency rating system for
homes may also affect incentives when a home is just above or below a
quality threshold. Similarly, we find evidence that builders are
responding to thresholds in the green housing market.
B. AEGB Program
The city of Austin started the Austin Energy Star program in 1985
as one of many energy conservation measures designed to conserve energy
so that Austin could avoid building a new power plant. This program is
one of the few green building rating programs to predate LEED, which
began in 1991. The Austin Energy Star program eventually evolved into
AEGB and changed its name when it became part of the municipal utility,
Austin Energy.
Austin Energy has been proactive in encouraging energy efficiency,
through AEGB and additional city ordinances. Homes in Austin are
required by city code to undergo an energy audit before they are sold.
However, there are many exemptions to this policy including exemptions
for homes less than 10 years old and homes that have performed energy
efficient upgrades or have received weatherization assistance. Since we
limit our analysis to homes that are less than 15 years old, most of the
homes in our sample would be exempt. If undertaken, these audits provide
information on the condition and effectiveness of attic insulation, air
leakage from the duct system, exterior doors, plumbing penetrations and
attics, age, efficiency and overall condition of heating and cooling
equipment, total square feet and direction of windows receiving more
than 1 hour of direct sunlight a day, and opportunities to improve
energy efficiency. These audits neither require sellers to make energy
efficiency improvements nor do they rank or categorize homes in terms of
greenness. While these audits may allow homebuyers to be aware of the
energy usage of their potential home, understanding the energy
implications of all this information may still be difficult for the
homebuyer as there is no summary rating. The audits are no guarantee
that homes are green and meet certain energy efficiency thresholds.
AEGB rates the greenness of homes on a 1 -5-star system that takes
into account five environmental impact topics: energy usage, water
usage, material and products, indoor air quality, and community. The
details have changed slightly over time, but the program has remained
similar in spirit. The 1-star rating is the "entry level" and
indicates that a home has fulfilled the basic requirements of the AEGB
program. (2) Once a home has achieved the basic requirements, it may
receive additional points for green features. If a home has more than 50
points it will receive an additional star. This cutoff was 60 points in
earlier versions (see Table 1). (3) Points may be obtained from adopting
green building practices approved by AEGB that may improve indoor air
quality, including items such as the following examples: "flooring
is 100% durable material" (4 points), enhance energy efficiency
"ceiling fans in all bedrooms" (2 points), encourage green
behavior "dedicated kitchen recycling center" (1 point) or
"attend AEGB Green by Design workshop" (1 point), mitigate
construction waste "on-site facilitation of sorting and reuse of
scrap building material" (2 points) or any other approved dimension
in the AEGB environmental impact topics. In addition to meeting the
requisite number of points, each additional star above two comes with
mandatory requirements. These requirements are cumulative. Thus, the
5-star home would need to meet all basic requirements plus 2-, 3-, and
4-star requirements as well as 5-star-specific requirements. (4) AEGB
reports that builders who seek a three or higher rating generally
incorporate the rating as a selling point but that many 1- and 2-star
homes may not even be aware they are rated. (5) Homes that achieve 4- or
5-star ratings are given plaques that are often prominently displayed at
open houses.
Note that the AEGB ratings combine private benefits relating to
reduced energy costs as well as public benefits in terms of reduced
construction waste and fewer environmental emissions. Consumers
potentially could value both the private energy cost savings as well as
the external public benefits. The public-private composition of the
ratings is not available so that it is not feasible to distinguish the
price-premium for different classes of benefit effects.
The city of Austin has supported the AEGB program through zoning
requirements for redeveloped areas, density bonuses for rated buildings,
and municipal programs such as the S.M.A.R.T. (Safe, Mixed-income,
Accessible, Reasonably-priced, Transit-oriented) program that provides
affordable housing for low- and moderate-income residents of Austin. All
S.M.A.R.T. homes are required to have at least a 1-star rating. Of the
9,943 homes ever rated by AEGB, 4,533 are S.M.A.R.T. homes, and 75% of
these homes have 1 star. In 2003, the City Council required all downtown
construction to earn a green building rating and in 2004, the City
Council adopted the Master Development Agreement for the Robert Mueller
Municipal Airport Redevelopment. The Mueller Redevelopment project
involved turning the old airport location into an urban village where
every home was required to have at least 3 AEGB stars. This location has
been described as the hottest real-estate area in the city as it is
close to downtown but with many new homes available and has easy access
to shopping and green space (e.g., Novak 2012). The existence of these
regulatory requirements will affect the distribution and possible price
premium associated with these ratings. Ignoring these regulatory
requirements will result in biased estimates for ratings. Our analysis
includes controls for the S.M.A.R.T. program and the Mueller urban
village redevelopment. We do not analyze the impact of the downtown
requirement since it primarily applies to commercial properties and only
two downtown properties were sold in our data period, both of which were
built before the requirement.
C. Energy Star Homes
The Energy Star program, which is administered by the U.S.
Environmental Protection Agency, began rating new homes in 1995. Unlike
AEGB, which evaluates a home on many aspects of green, Energy Star Homes
is based on the Home Energy Rating Systems (HERS). HERS has been
described as "an mpg for houses" and is calculated by
comparing the home to be rated against a similar home of the same size
and shape that meets the 2006 International Energy Conservation Code. A
HERS index of over 100 is a home that is less efficient than the
comparison home, while a lower HERS index is a home that is more
efficient. Each one point decrease in the HERS index indicates a 1%
improvement in energy efficiency relative to the reference home. Energy
Star Homes requires that a home be built with tight ducts and have a
HERS index of less than 85 so that Energy Star Homes are at least 15%
more efficient than a typical comparable home. (6)
Energy Star Homes come with a blue label, typically affixed to the
circuit breaker box, indicating they are an Energy Star Home. These
labels do not include the HERS index. The homeowners may be informed of
their HERS index through an Energy Star certificate or other
documentation, but there is no available data on the extent to which
consumers have received and understood the overall HERS index score.
There is evidence that consumers systematically misunderstand mpg
(Larrick and Soil 2008), which is arguably simpler than the HERS index.
Energy Star simplifies the HERS index into a yes or no rating, thus
acting as a green seal of approval which does not distinguish between
"efficient" and "very efficient" homes in a less
refined way than AEGB distinguishes between different levels of
"green" homes.
D. Environments for Living
The EFL is a private initiative of Masco Home Services. EFL
certifies homes based on performance requirements such as heating and
cooling efficiency, air tightness, and moisture control. EFL works with
the builder to approve plans and test the home and will arrange Energy
Star certification for an additional fee. EFL standards are more
stringent than Energy Star and, as a result, one would expect homes
meeting these standards to be more expensive. An EFL certification
provides a unique 2-year guarantee that the home will not use more than
a certain amount of energy and that the temperature of each room will
not vary more than 3[degrees]F from the thermostat.
All three rating programs in this study use evaluative ratings, but
the structure of the ratings differs with respect to whether they
provide green quality ratings or simply a green seal of approval.
Furthermore, Energy Star Homes focuses only energy efficiency while AEGB
and EFL ratings reflect many green quality attributes some of which have
private benefits (improved indoor air quality), public benefits (lower
construction waste), and benefits that are a mix of public and private
(energy efficiency) into a single overall green indicator. An AEGB 1
star or greater home and any EFL home will qualify as an Energy Star
Home, but 1 star and EFL homes will have additional green features such
as low VOC paints and water efficient appliances, depending on the
rating version used. Many homes have multiple ratings, such as an AEGB
star and an Energy Star certification.
Each of the rating systems active in Austin functions primarily as
a method to reduce highly technical information on the effectiveness of
a building's envelope, insulation R-scores, HVAC capacity, roofing
solar reflectiveness, water heating efficiency, and other aspects of
green to a simpler score whether it be a 1-5 star rating or a binary
rating such as the Energy Star seal of approval. While some of this
information is available via home inspections and home audits,
homebuyers may not possess the required knowledge to ask about and
interpret these measurements, especially for goods as heterogeneous as
housing. Thus, while homebuyers may value higher levels of green, they
may not be able to observe differences in green or have confidence that
a particular type of HVAC is the greenest choice for their house,
without the help of a rating system. We show this phenomenon in Figure 1
where we posit total costs and total benefit curves for a particular
segment of the market. We measure greenness along the horizontal axis in
terms of points achieved (p) through an AEGB program, but the concept
applies to other measures such as energy efficiency or a house's
environmental footprint. Total benefits TB(p) are concave and increasing
functions of p, and total costs TC(p) are convex and increasing. As the
lemons model suggests, homebuyers are only willing to pay for the
minimum greenness of a particular monitorable threshold. Thus, the price
premium for points p is not the continuous function TB(p) but rather is
a step function Price Premium)(p) reflecting the discontinuous
incentives from the evaluative categories as the points cross the
quality thresholds. A point that does not cross a quality threshold does
not lead to a price premium, even if it does benefit the homebuyer,
because the additional benefit within the ratings category cannot be
distinguished. Thus while [p.sup.*] may be the optimal amount of points
in terms of producer and consumer welfare in that it generates the
greatest spread between TB(p) and TC(p), the builder will only build a
home to either the 3-star threshold ([t.sub.3]) or the 4-star threshold
([t.sub.4]). Governmental advocates of green building techniques could
manipulate the thresholds to maximize the greenness of a particular
community:
(1) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where [g.sub.i] is the level of green chosen by home builder in
market segment i, [t.sub.1], [t.sub.2], [t.sub.3], [t.sub.4], [t.sub.5]
are the thresholds for each star rating, [theta] is a set of demand and
cost parameters, and f is the density function of market segments.
However, cost considerations also enter in any efficiency-based green
policy-ratings.
Manipulation of thresholds would only make sense if builders
respond to the price incentives provided by the rating system and if
additional points are not valued by homeowners except when they increase
the price of a home beyond a particular threshold. Because different
market segments will derive different benefits from greenness, we would
also expect to see different price premiums across market segments. In
the next section, we describe our data, then turn to showing builders
strategically respond to the incentives of a particular rating system in
which homebuyers are willing to pay for green ratings but not for
additional greenness not reflected in the ratings. We also assess
variation in price premiums for green labels across the market segments.
III. HOUSING AND ENERGY RATING DATA
To create the dataset of home ratings, housing prices, and housing
characteristics we merged two datasets. The first dataset comes from
Austin Energy and lists the addresses, star ratings, point totals, and
rating version for 9,951 homes that have been rated from between
February 1995 and January 2012. The second dataset is from the Austin
Board of REALTORS[R] (ABoR) and contains 30,515 observations on houses
that were sold through the Multiple Listing Service (MLS) between
January 2008 and April 2012. (7) The data from ABoR contain information
on sale price, sale date, housing characteristics, and addresses for
houses. Based on street address and unit number, we matched rated homes
to sales data resulting in an overlap of 1,105 homes that are rated by
AEGB and sold between January 2008 and April 2012 and 29,416 unrated
control houses sold during the same time period. We restrict the sample
to buildings built after the year AEGB began rating houses, 1995. This
reduces the sample to 15,668 homes and an overlap of 1,079 AEGB homes
that were rated and sold.
[FIGURE 1 OMITTED]
The ABoR includes a field that tells if the home was rated by any
green rating program. These ratings include AEGB, Energy Star Homes, and
EFL. There are other green ratings such as LEED and the National
Association of Home Builders ratings, but there are so few of these
homes in the sample that we do not analyze these programs. The ABoR data
tell us whether the listing agent reported that the home was rated by
AEGB (but not the rating). Because we have matched the ABoR data with
the AEGB rating, we can see how accurately the realtors report green
ratings. Table 2 compares the number of homes that were reported to have
an AEGB rating by the agent versus the number of homes that had a rating
based on matching between the AEGB and ABoR database. Many real estate
agents failed to report that AEGB rated the home, however, a much higher
percentage of 5-star ratings were reported (84%) versus 1-star ratings
(1.2%). (8) This result for the real-estate agents is consistent with
the experience of AEGB employees that 4- and 5-star home owners are
generally aware of their green status while 1- and 2-star homes are not.
Furthermore, it generalizes a finding in the housing market that sellers
of housing units with high quality features are more likely to disclose
that information on the MLS (Carrillo, Cellini, and Green 2013).
Table 3 presents summary statistics for rated buildings with sales
data and the control group of buildings. Rated buildings sell for less
on average than unrated buildings. A principal contributor to this
result is the prevalence of S.M.A.R.T. homes. If we look at rated homes
that are not part of the S.M.A.R.T. program, the mean sales price is
$314,981, which is higher than the price for unrated homes and rated
homes on average. Rated buildings are generally newer than unrated
buildings, reflecting the fact that most ratings are for new
construction projects, and green building has been one of the few
growing segments of the construction industry.
The size of rated and unrated homes, and the number of bedrooms and
the number of bathrooms are similar, but rated homes are less likely to
have a pool. Rated homes are more likely to be foreclosed upon, but this
is largely because of the high participation of the S.M.A.R.T. program.
Rated homes are much newer than unrated homes, with 65% of rated homes
having been built in the last 6 years. In general, rated homes and
unrated homes tend to be sold evenly throughout the years in the
dataset. The reason for there being a lower percentage of homes sold in
2012 is because the sales data end in April 2012.
Table 4 presents sample summary statistics by rating. Most homes
are not rated, and of the homes that are AEGB-rated, 60% have the 1-star
rating. Most of those 1-star ratings (74%) are also S.M.A.R.T. homes. To
capture homes built as part of new subdivisions, we created a dummy
variable that is equal to one if the builder of a home built more than
100 homes in the AEGB database (the ABoR data did not indicate builder).
In the full AEGB database, 2,924 homes were built by six companies. Of
these homes, 363 of them were also found in the sales data. The majority
of these homes (95%) are 1-star homes, and 54% of 1-star homes were
built by one of these six builders. With the exception of 1-star homes,
in the summary statistics more stars are associated with higher values
of homes, as we document below.
IV. BUILDER RESPONSES TO GREEN LABELS
The presence of green ratings will affect consumer demand for
homes, which in turn will alter the incentives builders have to offer
houses with green characteristics. Each AEGB-rated home meets certain
basic requirements and then achieves points based on green features. (9)
Because consumers monitor the star rating and not the overall points
score, the price premium should follow the step function pattern in
Figure 1. We investigate whether the star thresholds serve as a target
for builders seeking to reap the price premium from moving up to a
higher star category by testing whether there is a statistically
significant number of homes just to the right of each star cutoff or
rating notch. Figure 2 shows the number of homes in each point total for
all homes rated under AEGB ratings Version 2008.2 and vertical lines
that correspond to each rating notch under Version 2008.2. There is a
substantial concentration of points just beyond the cutoff for each
particular star rating of three or more stars.
We would like to compare homes across versions, but the points
required for each star rating are not consistent. Table 1 summarizes the
point notches for each rating for which we have information. We create a
normalized point total (Z,) which allows us to combine data from these
different versions:
(2) [Z.sub.i], - [(Total [Points.sub.i] - [Notch.sub.s,v]) /
([Notch.sub.s+1,v] - [Notch.sub.s,v])] + s,
where s denotes the number of stars home i has and v denotes the
version of each rating notch. [Notch.sub.s,v] is thus the number of
points needed to achieve star s under rating version v and Total Points,
is the number of points home i has. The normalized point total measures
the percentage of the way each home is to obtaining the next star. For
instance, if a home is halfway between the 4- and 5-star rating, it
would have a normalized point total of [Z.sub.i], = 4.5. There were some
errors in the data. For instance, some homes were listed as having 3
stars but only had enough points for a 2-star rating. These observations
were treated as having missing point totals. Figures 3A and 3B present
histograms of the normalized point totals. Figure 3A shows all rated
homes with known notches and point totals, whereas Figure 3B only shows
those homes with a 3-star or higher rating. In both figures, there is
clear bunching to the right of 3-, 4-, and 5-star rating notches.
To quantify this bunching we use a similar strategy from Chetty et
al. (2011) and Ramnath (2013) to estimate a counterfactual density
function of normalized point totals. To estimate this counterfactual we
use the regression:
(3) [C.sub.j] = [q.summation over
(i=0)][[beta].sub.i][Z.sup.i.sub.j] + [5.summation over
(s=3)][R.summation over (k=s)][[PHI].sub.k][D.sub.j] +
[[epsilon].sub.j],
where [C.sub.j] is the number of homes in each normalized point bin
j, [D.sub.j] is an indicator variable for each bin found to the left of
a star threshold, q is the order of the polynomial used, and R is equal
to s plus the number of bins where bunching is assumed to have taken
place. We first estimate [[??].sub.i] and [[??].sub.k] then calculate
counterfactual density, ([[??].sup.cf.sub.j]), using these estimates but
omitting the contribution of dummy variables around the notch. Thus
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. The number of homes
located to the right of the notch in this counterfactual is
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII][less than or equal
to]. However, [[??].sup.0.sub.N] is an overestimate of the number of
homes that located to achieve an extra point because it does not account
that the homes around the notch came from other areas of the
distribution. To take this into account, we shift the counterfactual
density up to satisfy the integration constraint and define the
counterfactual as [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
using the values of [[??].sub.j] that are fitted from the regression:
(4) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
where L is the set of bins that are not directly to the right of
the notches and [[??].sub.N] = [5.summation over (s=3)][R.summation over
(k=s)][[PHI].sub.k][D.sub.j]. Since [[??.sub.N] is a function of
[[??].sub.i], we iterate Equation (4) using the estimated
[[beta].sub.i], until we reach a fixed point. We bootstrap standard
errors using the same procedure as Chetty et al. (2011). Using a bin
size of 0.01 normalized points, an 8th order polynomial, and creating
notches that are 0.5 normalized points wide to the right of the 3-, 4-,
and 5-star notches (where we see the most bunching), we find the excess
number of homes just to the right of the notches, [[??].sub.N], is 1,301
which is significantly different from 0 with a 95% confidence interval
of (198, 2,980).
[FIGURE 2 OMITTED]
This result suggests that the star thresholds served as a target
for builders seeking to reap the price premium from moving up to a
higher AEGB star category. The concentration of point scores just above
the quality level cutoffs is consistent with builders targeting efforts
to earn a higher green rating. This effect is new to the green energy
labeling literature pertaining to houses but is similar in spirit to the
finding by Sallee and Slemrod (2012) with respect to automobile
companies' efforts to increase the mpg ratings if they were just
below a critical threshold. Furthermore, voluntary labels do more than
just inform consumers. AEGB's star labels increase the supply of
green buildings and thus may improve environmental outcomes and reduce
energy consumption.
[FIGURE 3 OMITTED]
V. ESTIMATES OF THE HEDONIC HOUSE PRICE MODEL
To estimate the value of green home ratings, we use a conventional
hedonic model that controls for a large number of variables that
influence real estate prices including age, size, location, and
amenities (Rosen 1974). A positive value of green home amenities
simultaneously reflects the greater willingness of buyers to purchase
greener homes and the greater marginal costs associated with providing
greener homes. The hedonic model can be used to estimate the price
premium from having an Energy Star, Environment for Living, or AEGB 1-,
2-, 3-, 4-, or 5-star home. We first use a Box-Cox model to test among
functional forms and find the most empirical support for a log-linear
model. (10) The basic model we estimate is:
(5) [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII].
Below we explore variants of this approach using spatial error and
spatial lag models.
The logarithm of the sale price of a home ([Price.sub.it]) is a
function of five dummy variables ([star.sub.s,i]) to indicate whether a
house has 1, 2, 3, 4, or 5 stars, a dummy variable for whether the house
is an Energy Star Home (estar,), a dummy variable for whether the home
comes with an EFL guarantee ([EFL.sub.i]), house characteristics
([X.sub.i]), and fixed effects that control for the year sold
([[alpha].sub.y]), the month sold ([[omega].sub.m]), and each census
tract ([[delta].sub.c]). Standard errors ([[epsilon].sub.it]) were
clustered on the census tract. Because the first AEGB-rated new homes in
the sample are from 1995 and because Energy Star Homes was launched in
1995, we restricted the sample to homes built after 1995. To explore the
robustness of our model with respect to the influence of potential
endogeneity of price and green ratings, we also used a matching model
which is described below.
The house price regression estimates appear in Table 5. The first
green labeling measure we consider in column 1 is an overall indicator
variable of whether the house has achieved an energy efficiency rating
under any of the three programs--AEGB, Energy Star, or EFL. This generic
green rating variable captures the average effect of the three program
ratings and indicates a significant positive price effect. The various
household characteristic variables are stable across all five
regressions. The number of bedrooms has a negative sign, which is
puzzling, but it is correlated with square feet. When the square feet
variable is omitted, the coefficient for the number of bedrooms is
positive and significant. Foreclosures sell for substantially less,
pools increase the value of the property, and older homes are worth less
than newer homes.
There are significant positive price effects for two of the three
different green label programs (column 2). Energy Star ratings generate
no price premium presumably because the AEGB ratings and EFL rating are
more comprehensive. This result is robust across all the regressions.
However when we used zip code fixed effects, instead of census block
fixed effects, we found that Energy Star Homes did receive a price
premium. (11)
In column 3, we use dummy variables for each of the AEGB stars and
find that houses with 2 and 3 stars do not command statistically
significant price premiums but houses with 1-, 4-, and 5-star ratings
do. A 5-star home sells for approximately 21% more than a similar
unrated home (using the Halvorsen-Palmquist adjustment). Four-star homes
sell for approximately 19% more while EFL homes are worth 8% more. Using
the mean value of a home built after 1995, these effects translate to
approximately $15,000-$55,000 for the different levels of green
certification. The existence of a premium is similar to the findings in
commercial real estate that a LEED Platinum rating increases property by
67%, but the relative magnitude of the effect is much smaller.
Because of their relationship to requirements that housing meet
energy efficiency standards, the policy variables are also of interest.
S.M.A.R.T. homes and homes built by big builders sell for less than
other AEGB-rated homes. Homes in the Mueller urban village (where homes
are required to have 3 stars and higher) receive a large statistically
significant price premium even after controlling for the star rating.
After excluding the policy variables (column 4), we see that ignoring
the correlation of green ratings with the S.M.A.R.T. program and big
builders causes the 1-star rated homes to have a negative sign (although
not significant). This result leads us to believe that studies on the
impact of green ratings on real estate may be biased if they do not
account for local incentive programs that encourage low income housing
to achieve green ratings (e.g., the S.M.A.R.T. program) or require green
homes for the most desirable new development areas (e.g., Mueller urban
village). Using zip code level or even census block controls would not
have eliminated this problem since many mixed use development projects
may require green buildings for subsidized housing (such as S.M.A.R.T.)
but not for other projects in the same development.
Three star AEGB homes may be valuable, but the 3-star effect is
difficult to discern from the data because 117 of the 172 (68%) of the
3-star homes in our sample are in the Mueller urban village. It is still
curious why a 1-star rating is valuable, but not 2-star homes. One
reason may be the 1-star homes are found predominantly in the lower end
of the market (Table 4) where any sign of green features or energy
efficiency can distinguish these homes from other lower priced homes.
Two star homes are more likely to be found in higher end market segments
(Table 4) where a 2-star rating may be interpreted as "below
average." We will return to this result when we analyze the
quantile regressions. Some marketing to clarify the quality content of
the 2-star AEGB rating may alleviate this problem. (12)
Finally, column 5 of Table 5 presents results that are similar to
column 3 but include a measure of the extra points obtained by the
builder. The sample size is smaller because the sample is limited to
homes with point totals and for which the rating version point cutoffs
are known. The coefficient for additional points is not statistically
different from zero. There is no price premium for homes going beyond
point thresholds, consistent with Figure 1.
A. Propensity Score Matching Models
To address the potential codetermination of price and green
ratings, we run a robustness check using a propensity score matching
model to estimate the effect of a generic green label on housing prices.
Specifically, we use a probit to estimate the probability a home will
receive a rating, conditional on other covariates (the propensity score)
and then compared rated homes with unrated homes that had a similar
propensity score. We used a nearest-neighbor one-to-one matching and a
bandwidth of 0.06 as described in Leuven and Sianesi (2003). Table 6
presents these estimates. The propensity score method does not allow for
us to control for the price impacts of important variables such as the
S.M.A.R.T. program or being located in Mueller since all S.M.A.R.T. and
all Mueller homes are also rated. Thus, when we examine the impact of a
green rating without controlling for the S.M.A.R.T. program, we find the
rating comes with a negative price premium. Restricting the sample in
column 2 of Table 6 to homes that are not in the S.M.A.R.T. program, we
find that the premium is actually larger than what was found in column 1
of Table 5. Columns 3 and 4 use a linear regression framework to
estimate the effects for the homes that were matched in the propensity
score matching technique and find similar results to Table 5, when the
policy variables are taken into account.
B. Spatial Econometric Models
Spatial autocorrelation can cause OLS to be biased unless one
employs spatial econometrics techniques. There are two general varieties
of spatial econometric models, the spatial autoregressive model (spatial
lag) and the spatial error model. The spatial lag model implies the
price of a home is directly influenced by prices of nearby homes:
(6) y = [rho]Wy + X[beta] + [epsilon]
where [rho] is a scalar spatial autocorrelation parameter and W is
a spatial weight matrix. In a situation where the spatial lags are
present, results from OLS will be biased and inconsistent due to the
endogeneity of [rho]Wy.
The spatial error model posits that the spatial autocorrelation is
in the error term. This model can be represented as:
(7) y = X[beta] + u and
(8) u = [lambda]Wu + [epsilon]
where W is a weighting matrix as before and [lambda] is a scalar
spatial autocorrelation parameter. The spatial error model is
appropriate when there are omitted variables with a spatial component
that results in residual spatial error correlation. The green sentiment
of a neighborhood or a sense of community are examples of potential
omitted variables that are difficult to measure but may affect property
values. Kuminoff, Parmeter, and Pope (2010) found spatial regression
techniques substantially reduce bias from cross-sectional regression
techniques. In the spatial error model, the OLS coefficients are
unbiased but inconsistent.
The results of the spatial lag and spatial error models are
presented in Table 7. Interpreting the results of the spatial error
model is straightforward. However, with the spatial error model, the
total effect is decomposed into a direct effect, [beta], and an indirect
effect, (1/1 - [rho]). The results qualitatively are similar to Table 5,
but the magnitude of the 1-star premium is more robust in the spatial
regressions. The premium for 4- and 5-star ratings remain strongly
significant and large. The discount for being in the S.M.A.R.T. program
is lower with the spatial models, and being in the Mueller redevelopment
district is less valuable in the spatial models.
VI. EFFECTS ACROSS THE HOUSING PRICE DISTRIBUTION
The role of green labels will vary across the distribution of
houses in the market. There are three principal influences at work.
First, the price premium commanded by a higher rated home will be
greater for homes that have greater energy efficiency characteristics
than other homes in their price range. Thus, the price premium could be
greater for lower priced homes if homes at the low end of the market are
less likely to have energy efficiency amenities in the absence of a
green rating. Second, we expect that the incremental cost of obtaining a
high quality certification may be lower at the upper end of the market
since one would expect higher quality homes to already have more
insulation and other characteristics that should reduce the cost of
achieving a particular energy efficiency rating. An offsetting
consideration is that higher end homes tend to be larger so that the
cost of achieving a level of energy efficiency and green status
generally will be greater. Finally, if a greener rating is viewed as an
environmental benefit rather than simply a cost saving measure, and if
there is a positive income elasticity of demand for environmental
benefits, then we would expect homeowners with more disposable income to
demand higher end homes that are also greener.
To examine these influences we first consider the distribution of
energy efficiency ratings across different quartiles of the market.
While most of the homes listed in Table 8 do not have energy efficiency
ratings, the distribution across the market of homes that are rated
accords with expectations. The 1-star AEGB ratings are concentrated
among homes below the median house value. In contrast, houses with
ratings of 2 stars and above are concentrated at the upper end of the
market. The regulatory requirements that houses in the Mueller
Redevelopment project have a rating of 3 or more stars in the AEGB
system contributes to this clustering. The especially large number of
lower end homes with a 1-star rating is the result of the regulatory
requirement that the S.M.A.R.T. program homes meet the AEGB 1-star
rating criteria. In the absence of these regulatory requirements, one
would have expected a distribution of AEGB ratings more similar to those
for Energy Star and EFL.
For the other energy efficiency ratings that are not influenced by
the regulatory requirements, the patterns are more consistent. In the
case of the Energy Star ratings and the EFL ratings, there is a dominant
representation of such ratings among the homes above the median house
price. This result is what one would expect given the relatively lower
cost of achieving this rating for the more expensive homes that already
incorporate more costly features.
While the prevalence of highly energy efficient homes is
concentrated at the upper end of the market, there are nevertheless
rewards for greater energy efficiency throughout the market. Table 9
presents the quantile regressions for the log of the housing price as a
function of the various energy ratings and house price characteristics.
For the AEGB ratings, the 1 -star ratings yield significant price
dividends in all but the most expensive homes where meeting a low end
energy efficiency standard may not be a positive quality signal. For
each of the quantile regressions, the largest effects are exhibited by
the 5-star ratings, which is what one would expect if the highest energy
efficiency rating is valued by the market. The Energy Star rated homes
do not receive a premium across any of the quantiles. The EFL rated
homes have a consistent significant positive price effect across the
market.
Together these sets of results suggest that ratings of higher
energy efficiency do command a market premium, and this premium affects
all segments of the market. However, due to the substantial incremental
cost of achieving the energy efficiency ratings and a positive income
elasticity of demand for environmental goods, the more demanding ratings
are more heavily concentrated among the upper end homes.
VII. CONCLUSION
Controlling for other housing characteristics, green ratings that
provide valuable quality information to consumers should generate price
premium. On average this prediction is borne out as the green ratings
used in the Austin, Texas market generate an average residential price
premium of 5%. The effective ratings programs were the Environments for
Living program, 1-star AEGB ratings, and the upper end of the AEGB
quality ratings. The 2- and 3-star AEGB rating was only valuable for
high quality homes. For ratings of 4 and 5 stars, there is evidence that
builders target house characteristics to achieve a higher rating score.
This bunching around housing notches appears to be a function of
producers strategically building homes to achieve ratings, which is
consistent with the absence of a price premium for points beyond ratings
cutoffs. Also consistent with economic models of quality certification
is that highly rated houses with 4-5 stars in the AEGB system command a
larger premium than lesser starred homes, with the point estimate of the
effect increasing with the number of stars in the rating.
The Energy Star Homes certification did not receive a positive
premium in any of our regressions, which is likely a function of it
being the least stringent of all the labels examined. Kok and Kahn
(Forthcoming) did find that Energy Star certification generated a
positive price premium in California. However, the difference in the
results may be because they used zip code-street name level fixed
effects instead of more refined, census tract fixed effects. They also
used non-spatial methods, whereas we used spatially explicit regression
techniques. (13) Furthermore, most of Kok and Kahn's regressions
examined the effect of a generic green label which included LEED homes
and a local energy rating. An industry study in Houston found that
Energy Star Homes did not result in significant energy savings compared
to other homes partially because all homes are becoming more efficient
(Hassel and Blasnik 2009; Holladay 2011). The threshold for Energy Star
Homes may simply be too low to attract a price premium.
While green ratings appear to be generally effective in achieving
their objective of influencing the supply of green homes and fostering a
price premium for energy efficient and environmentally friendly homes,
the presence of an effect does not necessarily imply that the programs
are entirely successful. Consumers may overestimate or underestimate the
underlying house characteristics associated with the ratings so that the
need for additional research to examine the relationship between
consumer beliefs, the rating scales, and the green characteristics of
houses is broader than just for the 1-star homes.
ABBREVIATIONS
AEGB: Austin Energy Green Building
ABoR: Austin Board of REALTORS[R]
EFL: Environments for Living
HERS: Home Energy Rating Systems
HVAC: Heating, Ventilation and Air Conditioning
IC3: International Code Compliance Calculator
LEED: Leadership in Energy and Environmental Design
MLS: Multiple Listing Service mpg: miles per gallon
S.M.A.R.T.: Safe, Mixed-Income, Accessible, Reasonably-Priced,
Transit-Oriented Program
USGBC: U.S. Green Building Council
VOC: Volatile Organic Compound
doi: 10.1111/ecin.12140
REFERENCES
Akerlof, G. A. "The Market for 'Lemons': Qualitative
Uncertainty and the Market Mechanism." Quarterly Journal of
Economics, 84, 1970, 488-500.
Aroul, R. R., and J. A. Hansz. "The Role of Dual-pane Windows
and Improvement Age in Explaining Residential Property Values." The
Journal of Sustainable Real Estate, 3(1), 2011, 142-61.
Blackman, A., and J. Rivera. "The Evidence Base for
Environmental and Socioeconomic Impacts of 'Sustainable'
Certification." Resources for the Future Discussion Paper 10-17,
2010.
Brouhle, K., and M. Khanna. "Information and the Provision of
Quality Differentiated Products." Economic Inquiry, 45(2), 2007,
377-94.
Brounen, D., and N. Kok. "On the Economics of Energy Labels in
the Housing Market." Journal of Environmental Economics and
Management, 62, 2011, 166-79.
Carrillo, P, S. Cellini, and R. Green. "School Quality and
Information Disclosure: Evidence from the Housing Market." Economic
Inquiry, 51(3), 2013, 1809-28.
Chegut, A., P. Eichholtz, and N. Kok. "Supply, Demand and the
Value of Green Buildings." RICS Research, 2012.
Chetty. R., J. N. Friedman, T. Olsen, and L. Pistaferri.
"Adjustment Costs, Firm Responses, and Micro vs. Macro Labor Supply
Elasticities: Evidence from Danish Tax Records." Quarterly Journal
of Economics, 126, 2011, 749-804.
Eichholtz, P., N. Kok, and J. M. Quigley. "Doing Well by Doing
Good? Green Office Buildings." American Economic Review, 100, 2010,
2492-509.
--. "The Economics of Green Building." Review of
Economics and Statistics, 95(1), 2013, 50-63.
Fuerst, F., and P. McAllister. "Green Noise or Green Value?
Measuring the Effects of Environmental Certification on Office
Values." Real Estate Economics, 39(1), 2011, 45-69.
Harrison, D., and M. Seiler. "The Political Economy of Green
Building." Journal of Property Investment and Finance, 29(4-5),
2011,551-65.
Hassel, S., and M. Blasnik. "Houston Home Energy Efficiency
Study," Advanced Energy, 2009. Accessed April 2, 2014.
http://www.resnet.us/blog/wp-content/
uploads/2012/08/Houston-Energy-Efficiency-Study2009-Final.pdf.
Hibbard, J. H., and E. Peters. "Supporting Informed Consumer
Health Care Decisions: Data Presentation Approaches that Facilitate the
Use of Information in Choice." Annual Review of Public Health, 24,
2003, 413-33.
Hibbard, J. H., J. Greene, S. Sofaer, K. Firminger, and J. Hirsh.
"An Experiment Shows That a Well-Designed Report on Costs and
Quality Can Help Consumers Choose High-Value Health Care." Health
Affairs, 31 (3), 2012,560-68.
Holladay, M. "Disappointing Energy Savings for Energy
Efficient Homes." Green Building Advisor, 2011. Accessed April 2,
2014. http://www.greenbuildingadvisor.com/
blogs/dept/musings/disappointing-energy-savingsenergy-star-homes.
Hotz, V. J., and M. Xiao. "Strategic Information Disclosure:
The Case of Multiattribute Products with Heterogeneous Consumers."
Economic Inquiry, 51(1), 2013, 865-81.
Johnson, R., and D. Kaserman. "Housing Market Capitalization
of Energy-Saving Durable Good Investments." Economic Inquiry,
21(3), 1983, 374-86.
Kok, N., and M. E. Kahn. Forthcoming. "The Capitalization of
Green Labels in the California Housing Market." Regional Science
and Urban Economics.
Kuminoff, N. V., C. F. Parmeter, and J. C. Pope. "Which
Hedonic Models Can We Trust to Recover the Marginal Willingness to Pay
for Environmental Amenities?" Journal of Environmental Economics
and Management, 60(3), 2010, 322-35.
Lande, C. D. "Homeowner Views on Housing Market Valuation of
Energy Efficiency." M.A. Dissertation, University of Montana, 2006.
Larrick, R. P., and J. B. Soli. "The MPG Illusion."
Science, 320(5883), 2008, 1593-94.
Leuven, E., and B. Sianesi. "PSMATCH2: Stata Module to Perform
Full Mahalanobis and Propensity Score Matching, Common Support Graphing,
and Covariate Imbalance Testing," 2003. Accessed August 5, 2014.
http://ideas.repec.Org/c/boc/bocode/s432001.html.
Nevin, R., and G. Watson. "Evidence of Rational Market
Valuation for Home Energy Efficiency." The Appraisal Journal, 4,
1998, 401-9.
Novak, S. "Mueller Set for Quantum Leap in Growth this
Year." Austin America-Statesman, 2012. Accessed May 8,2013.
http://www.statesman.com/news/business/realestate/mueller-set-for-quantum-leap-in-growth-thisyear-l/nRkmn/.
Ramnath, S. "Taxpayers' Responses to Tax-Based Incentives
for Retirement Savings: Evidence from the Saver's Credit Notch
Journal of Public Economics, 101, 2013, 77-93.
Rosen, S. "Hedonic Prices and Implicit Markets: Product
Differentiation in Pure Competition." Journal of Political Economy,
82(1), 1974, 34-55.
Sallee, J. M., and J. Slemrod. "Car Notches: Strategic
Automaker Responses to Fuel Economy." Journal of Public Economics,
96(11-12), 2012, 981-99.
Teisl, M. F., J. Rubin, and C. L. Noblet. "Non-Dirty Dancing?
Interactions Between Eco-Labels and Consumers." Journal of Economic
Psychology, 29, 2008, 140-59.
Viscusi, W. K. "A Note on 'Lemons' Markets with
Quality Certification." Bell Journal of Economics, 9, 1978, 277-79.
(1.) These costs are modest for AEGB ($50), but higher for Energy
Star Homes and EFL depending on who does the certification.
(2.) The basic requirements change from year to year to keep pace
with technological advancements in energy efficiency and green building
techniques. The current version, 2013.0, has multiple basic
requirements. It requires homes to have a minimum International Code
Compliance Calculator (IC3), an energy code optimized for climate zone
3, score of 0.0 or a Home Energy Rating Score (HERS) of 70, indicating
this home is 30% more efficient than the average home built to code.
Homes must meet appropriate heating, ventilation and air conditioning
(HVAC) efficiency and design targets, have appropriate insulation with
no added urea formaldehyde and meeting other technical targets, have no
skylights in conditioned space or have skylights that meet Energy Star
guidelines, have exhaust fans for cooktops and bathrooms where the
bathroom exhaust fans are connected to timers or humidistat,
low-volatile organic compound (VOC) interior wall and ceiling paint or
City of Austin recycled paint, installed carbon monoxide detectors, at
least four Energy Star appliances, ventilation fans, light fixtures or
ceiling fans, a minimum of two ceiling fans installed within heated and
cooled spaces, planting beds with a minimum depth of 6" of soil
containing 25% compost and a minimum of 2" organic mulch, and a
minimum of 90% of new plants are from Grow Green plant list.
(3.) The most recent version, version 2013.0 has a 1-star threshold
of 25 points.
(4.) Version 2013.0 5-star requirements are meeting roofing solar
reflectance and roof slope standards, water heater efficiency standards,
and having certain energy management controls and monitoring.
(5.) Our analysis is based on matching addresses between all homes
that were ever rated and homes sold during the sample period. Thus, we
may be aware of the rating even if the homeowner and realtor are not.
(6.) Homes in climate zones 6-8 are required to have a HERS index
lower than 80 for Energy Star; however Austin, TX is zone 2.2. Energy
Star for Homes Versions 2.5 and 3 have variable energy efficiency
targets but incorporate indoor air quality concerns and require a
moisture control plan. We assume all the homes in our database are
certified under Energy Star for Homes Version 2 or older.
(7.) The MLS only includes properties sold by a realtor with access
to the MLS and thus does not include nonarms-length transactions. The
sample does not include enough repeat sales to perform a meaningful
analysis of a repeat sales sample.
(8.) With the exception of the analysis for Table 2, when we refer
to AEGB stars these are the actual star rating achieved by the home,
regardless of whether ABoR reported the rating.
(9.) We focus our analysis of producer response to AEGB labels
since we only have data on the point totals from the AEGB labels. There
is likely a similar response with bunching around the HERS rating for
Energy Star or other green features with EFL.
(10.) Using census tract level fixed effects we estimate [theta] =
0.51 *** and [lambda] = 1.33 ***. We reject the null hypothesis that
[theta] = 0, [theta] = 1, [lambda] = 0, and [lambda] = 1, which suggests
a general [theta] and X. However, the estimates based on general
[lambda] and [theta] models are difficult to interpret. Following
Cameron and Trivedi (2009, pg. 95), we find that our values for [theta]
and [lambda] suggest a log-linear model.
(11.) The p value for the difference between the average AEGB
rating and the EFL rating is 0.39. All other p values (difference
between AEGB and Energy Star, EFL and Energy Star, and whether all three
are equal to each other) are less than 0.01.
(12.) The stigmatization of lower certified homes is not a problem
with the LEED program, which ranks homes as Certified, Silver, Gold, or
Platinum. These categories all imply positive quality characteristics
and are less likely to be interpreted as "below average."
(13.) Indeed, a previous version of our paper found significant
premiums for Energy Star Homes when we used zip code level fixed
effects.
SHARON SHEWMAKE and W. KIP VISCUSI *
* We are grateful for the help of Bryan Bomer at Austin Energy and
Stan Martin at the Austin Board of REALTORS[R] for providing the data
used in this analysis and for explanations of the green building market
in Austin, Texas.
Shewmake: Assistant Professor, Department of Economics, Western
Washington University, Bellingham, WA 98225. Phone 1-850-591-0202, Fax
1-360-650-3910, E-mail sharon.shewmake@wwu.edu
Viscusi: University Distinguished Professor of Law, Economics, and
Management, Department of Law, Economics, and Management, Vanderbilt
University, Nashville, TN 37203. Phone 1-615-343-7715, Fax
1-615-322-5953, E-mail kip.viscusi@vanderbilt.edu
TABLE 1
Point Notches for AEGB Green Home Rating
Versions
Homes
Rated under
2-Star 3-Star 4-Star 5-Star Version with
Version Notch Notch Notch Notch Point Totals
6/6.0 60 90 130 180 2,469
7/7.0 60 90 130 180 2,516
7.1 60 90 130 180 1,382
7.2 60 95 130 180 1,241
8/8.0 60 85 115 150 115
2008.2 50 75 100 125 636
Note: Rating versions not included because their point
notches are not known include versions II, 8.1, 8.2, 9.2,
2008.1, 2010.0, 2010.1, and 2010.2.
TABLE 2
Reported and Actual AEGB Ratings
Percent of ABoR
ABoR Listings Listings
that Report Total Number of Accurately
AEGB Rating in AEGB Homes in Reporting AEGB
Sample Sample Rating
1 Star 8 642 1.2
2 Star 3 184 1.6
3 Star 54 162 33.3
4 Star 17 35 48.6
5 Star 47 56 84.0
TABLE 3
Merged Sales and AEGB Data Summary
Statistics
AEGB Rated
Buildings Sold
between 2008 Control
and 2012 Buildings
Sample size 1,079 14,589
Sales price (in thousands 220.01 249.91
of dollars) (183.11) (238.00)
Dwelling size (thousands 1.95 2.27
of square feet) (0.66) (0.96)
Number of bedrooms 3.30 3.43
(0.68) (0.83)
Number of bathrooms 2.40 2.53
(0.54) (0.74)
Foreclosure (1 = yes) 0.21 0.16
Pool 0.01 0.07
Age (in years) 4.54 5.83
(3.75) (4.28)
Number Energy Star rated 60 255
Percent Energy Star rated 5.56 1.75
Age (22.93) (1.31)
0-5 years 0.65 0.52
6-10 years 0.26 0.31
11-16 years Year sold 0.08 0.18
2009 0.34 0.30
2010 0.31 0.29
2011 0.28 0.32
2012 0.08 0.09
Note: Standard deviations are in parentheses.
TABLE 4
Summary Statistics by AEGB Star Rating for Merged AEGB and Sales Data
No Stars 1 Star 2 Stars
Number of homes 14,589 642 184
Energy Star Homes 255 13 0
Environments for Living homes 93 0 0
S.M.A.R.T. homes 0 474 34
Mueller development 0 0 0
Big builder 0 344 6
Foreclosed homes 2,275 186 25
Average age of homes 5.83 4.45 8.10
Number of homes of age (4.28) (3.49) (3.12)
1 -2 Years 1,779 77 7
2-3 Years 567 30 4
3-4 Years 976 68 5
4-5 Years 1,310 103 11
5-6 Years 1,218 78 10
6-10 Years 3,652 144 73
10-15 Years 3,430 56 72
Average price of homes ($ thousands) 249.91 147.71 257.25
(238.0) (61.0) (120.9)
Number of homes in first price quantile 3,444 334 40
Number of homes in second price quantile 3,686 211 8
Number of homes in third price quantile 3,580 63 79
Number of homes in fourth price quantile 3,891 34 57
3 Stars 4 Stars 5 Stars
Number of homes 162 35 56
Energy Star Homes 35 3 9
Environments for Living homes 0 0 0
S.M.A.R.T. homes 25 1 11
Mueller development 117 15 19
Big builder 12 0 0
Foreclosed homes 14 0 1
Average age of homes 2.65 1.54 1.21
Number of homes of age (2.88) (2.77) (1.45)
1 -2 Years 65 10 13
2-3 Years 17 1 4
3-4 Years 17 5 7
4-5 Years 10 1 2
5-6 Years 5 0 0
6-10 Years 20 0 1
10-15 Years 6 2 0
Average price of homes ($ thousands) 292.43 425.82 588.28
(135.2) (152.5) (501.1)
Number of homes in first price quantile 27 1 6
Number of homes in second price quantile 17 0 7
Number of homes in third price quantile 14 4 1
Number of homes in fourth price quantile 104 30 42
Note: Standard deviations are in parentheses.
TABLE 5
Regressions of Log Price on House Characteristics for Houses
Built after 1995
Age (1) (2) (3)
Green rating (1=yes) 0.05 ***
(0.01)
Any AEGB 0.06 ***
(0.02)
Energy Star 0.01 0.00
(0.02) (0.02)
EFL 0.09 *** 0.09 ***
(0.02) (0.02)
1-Star rating 0.08 ***
(0.02)
2-Star rating 0.02
(0.02)
3-Star rating 0.05
(0.05)
4-Star rating 0.17 **
(0.07)
5-Star rating 0.19 ***
(0.06)
Additional points
S.M.A.R.T. program -0.11 ** -0.12 ** -0.13 **
(0.04) (0.04) (0.05)
Big builder -0.03 -0.04 -0.04
(0.04) (0.04) (0.04)
Mueller 0.27 *** 0.26 *** 0.24 ***
(0.05) (0.05) (0.06)
Square feet (thousands) 0.36 *** 0.36 *** 0.36 ***
(0.02) (0.02) (0.02)
Number of bathrooms 0.05 ** 0.05 ** 0.05 **
(0.02) (0.02) (0.02)
Number of bedrooms -0.04 *** -0.04 *** -0.04 ***
(0.01) (0.01) (0.01)
Foreclosure -0.30 *** -0.30 *** -0.30 ***
(0.02) (0.02) (0.02)
Pool 0.13 *** 0.13" * 0.13 ***
(0.02) (0.02) (0.02)
Age
1-2 Years -0.02 ** -0.03 ** -0.02 **
(0.01) (0.01) (0.01)
2-3 Years -0.02 -0.02 -0.02
(0.01) (0.01) (0.01)
3-4 Years -0.03 * -0.03 * -0.03 *
(0.02) (0.02) (0.02)
4-5 Years -0.05 ** -0.06 ** -0.05 **
(0.02) (0.02) (0.02)
5-6 Years -0.06 *** -0.06 *** -0.06 ***
(0.02) (0.02) (0.02)
6-10 Years -0.10 *** -0.10 *** -0.10"'
(0.01) (0.01) (0.02)
10-15 Years -0.13 *** -0.13 *** -0.13 ***
(0.02) (0.02) (0.02)
Constant 11.46 *** 11.46 *** 11.46 ***
(0.05) (0.05) (0.05)
Observations 15,668 15,668 15,668
[R.sup.2] .85 .85 .85
Age (4) (5)
Green rating (1=yes)
Any AEGB
Energy Star 0.01 0.00
(0.02) (0.02)
EFL 0.09 *** 0.10 ***
(0.02) (0.02)
1-Star rating -0.03 0.11 *
(0.04) (0.06)
2-Star rating -0.01 0.03
(0.04) (0.03)
3-Star rating 0.06 0.07 **
(0.06) (0.03)
4-Star rating 0.18 ** 0.12 **
(0.07) (0.06)
5-Star rating 0.19 ** 0.17 **
(0.09) (0.06)
Additional points -0.05
(0.06)
S.M.A.R.T. program -0.12 **
(0.05)
Big builder -0.04
(0.04)
Mueller 0.27 ***
(0.05)
Square feet (thousands) 0.36 *** 0.36 ***
(0.02) (0.02)
Number of bathrooms 0.05 ** 0.05 **
(0.02) (0.02)
Number of bedrooms -0.04 *** -0.04 ***
(0.01) (0.01)
Foreclosure -0.30 *** -0.30 ***
(0.02) (0.02)
Pool 0.13 *** 0.13 ***
(0.02) (0.02)
Age
1-2 Years -0.02 ** -0.02 **
(0.01) (0.01)
2-3 Years -0.02 -0.02
(0.01) (0.01)
3-4 Years -0.03 -0.03 ***
(0.02) (0.02)
4-5 Years -0.05 ** -0.05 **
(0.02) (0.02)
5-6 Years -0.06 *** -0.06 ***
(0.02) (0.02)
6-10 Years -0.09 *** -0.10 ***
(0.02) (0.02)
10-15 Years -0.12 *** -0.13 ***
(0.02) (0.02)
Constant 11.46 *** 11.46 ***
(0.05) (0.05)
Observations 15,668 15,488
[R.sup.2] .85 .85
Note: New Construction is the omitted category for age.
Dummy variables were included for property type (house,
condo, duplex, townhouse, manufactured home, mobile home, or
other), for each census tract, and for each year and month a
house was sold. Standard errors, in parentheses, are cluster
robust at the census tract level.
Significance at the 0.10, 0.05, and 0.01 levels is indicated
by *, **, and ***, respectively.
TABLE 6
Regression Results Robustness Checks with Propensity Score Matching
(2)
Propensity
(1) Score (3) (4)
Propensity Matched Linear Linear
Score No S.M.A.R.T. Regression Regression
Matched Homes on Support on Support
Green rating -0.04 ** 0.27 ** 0.01 0.05 ***
(0.02) (0.02) (0.02) (0.02)
S.M.A.R.T. -0.11 **
program (0.04)
Big Builder -0.04
(0.05)
Mueller 0.28 ***
(0.03)
Observations 13,025 12,437 13,025 13,025
Note: Dummy variables were included for property type
(house, condo, duplex, townhouse, manufactured home, mobile
home, or other), whether the house was a foreclosure,
whether the house had a pool, vintage groups, for each
census tract, and for each year and month a house was sold.
Additional controls included square feet, number of
bedrooms, number of bathrooms. Standard errors are in
parentheses, they are cluster robust for regressions 3 and
4.
Significance at the 0.10, 0.05, and 0.01 levels is indicated
by *, **, and ***, respectively.
TABLE 7
Regressions of Log Price on House Characteristics for Houses
Built After 1995 Using Spatial Lag and Spatial Error Regressions
(1) (2) (3) (4)
Spatial Spatial Spatial Spatial
Lag 3 Lag 5 Error 3 Error 5
Nearest Nearest Nearest Nearest
Neighbors Neighbors Neighbors Neighbors
Energy Star -0.002 -0.002 0.01 0.01
(0.01) (0.01) (0.01) (0.01)
EFL 0.06 *** 0.08 *** 0.06 *** 0.06 ***
(0.02) (0.01) (0.02) (0.02)
1-Star rating 0.05 *** 0.05 *** 0.05 ** 0.05 **
(0.02) (0.02) (0.02) (0.02)
2-Star rating 0.02 0.01 0.01 0.003
(0.02) (0.02) (0.02) (0.01)
3-Star rating 0.05 * 0.05 0.01 0.001
(0.03) (0.03) (0.03) (0.03)
4-Star rating 0.17 *** 0.18 *** 0.12 *** 0.014 ***
(0.04) (0.04) (0.04) (0.04)
5-Star rating 0.17 *** 0.20 *** 0.13 *** 0.016 ***
(0.03) (0.03) (0.03) (0.03)
S.M.A.R.T. program -0.09 *** -0.08 *** -0.05 ** -0.04 *
(0.02) (0.02) (0.02) (0.02)
Big builder -0.01 -0.02 -0.04 ** -0.04 **
(0.02) (0.02) (0.02) (0.02)
Mueller 0.09 * 0.06 0.20 ** 0.017
(0.05) (0.05) (0.09) (0.09)
Square feet (thousands) 0.25 *** 0.24 *** 0.28 ** 0.28 ***
(0.00) (0.00) (0.00) (0.00)
Number of bathrooms 0.04 *** 0.04 *** 0.04 *** 0.03 ***
(0.00) (0.00) (0.00) (0.00)
Number of bedrooms -0.02 *** -0.02 *** -0.02 *** -0.02 ***
(0.00) (0.00) (0.00) (0.00)
Foreclosure -0.25 *** -0.24 *** -0.23 *** -0.23 ***
(0.00) (0.01) (0.00) (0.00)
Pool 0.11 *** 0.11 *** 0.12 *** 0.12 ***
(0.01) (0.01) (0.01) (0.01)
Constant 6.91 *** 6.41 *** 12.45 *** 12.45 ***
(0.07) (0.08) (0.09) (0.10)
Lambda 0.576 *** 0.665 ***
Rho 0.428 *** 0.467 ***
Observations 15,273 15,273 15,273 15,273
Log-Likelihood 3,409.3 3,496.6 3,111.43 3,390.2
Note: Dummy variables were included for property type (house,
condo, duplex, townhouse, manufactured home, mobile home, or
other), vintage groups, for each census tract, and for each year
and month a house was sold. Asymptotic standard errors are in
parentheses.
Significance at the 0.10,0.05, and 0.01 levels is indicated by *,
**, and ***, respectively.
TABLE 8
Number of Rated Homes by Housing Price
Quartile
0-25th 25th-50th 50th-75th 75th-100th
Percentile Percentile Percentile Percentile
No AEGB 3,506 3,674 3,747 3,662
Rating
1 Star 337 211 61 33
2 Stars 40 8 84 52
3 Stars 28 16 20 98
4 Stars 1 0 4 30
5 Stars 6 9 1 42
Energy 7 57 130 121
Star Rated
EFL 0 3 33 57
TABLE 9
Quantile Regressions of Log Housing Price on House Characteristics
10th 25th 50th
Percentile Percentile Percentile
Energy Star 0.00 0.00 0.01
(0.02) (0.01) (0.01)
EFL 0.11 *** 0.08 *** 0.08 **
(0.03) (0.01) (0.01)
1 Star 0.09 *** 0.09 *** 0.05 **
(0.02) (0.02) (0.02)
2 Stars 0.00 0.00 0.02
(0.03) (0.02) (0.01)
3 Stars 0.07 0.06 0.03
(0.12) (0.04) (0.03)
4 Stars 0.10 ** 0.14 *** 0.09 **
(0.04) (0.04) (0.04)
5 Stars 0.27 *** 0.19 *** 0.10 ***
(0.06) (0.05) (0.04)
Pool 0.07 *** 0.10 *** 0.12 ***
(0.01) (0.01) (0.01)
S.M.A.R.T. -0.06 ** -0.10 *** -0.07 ***
(0.02) (0.02) (0.02)
Big builder -0.04 -0.03 *** -0.07 ***
(0.03) (0.02) (0.02)
Mueller urban village 0.28 *** 0.41 *** 0.46 ***
(0.13) (0.04) (0.15)
Square feet (thousands) 0.31 *** 0.35 *** 0.37 ***
(0.01) (0.00) (0.01)
Bathrooms 0.03 *** 0.02 ** 0.00
(0.01) (0.01) (0.01)
Bedrooms -0.03 *** -0.04 *** -0.03 ***
(0.01) (0.01) (0.00)
Foreclosure -0.29 *** -0.26 *** -0.25 ***
(0.02) (0.01) (0.00)
Constant 11.25 *** 11.37 *** 11.46 ***
(0.02) (0.02) (0.02)
75th 90th
Percentile Percentile
Energy Star 0.02 0.01
(0.01) (0.01)
EFL 0.07 *** 0.05 ***
(0.02) (0.02)
1 Star 0.05 ** 0.02
(0.02) (0.02)
2 Stars 0.00 0.03 *
(0.00) (0.02)
3 Stars -0.02 0.11 *
(0.04) (0.06)
4 Stars 0.06 0.14 **
(0.05) (0.16)
5 Stars 0.12 *** 0.28 ***
(0.04) (0.08)
Pool 0.12 *** 0.15 ***
(0.01) (0.02)
S.M.A.R.T. -0.13 *** -0.14 ***
(0.02) (0.04)
Big builder -0.04 ** -0.01
(0.02) (0.02)
Mueller urban village 0.33 *** 0.12
(0.08) (0.14)
Square feet (thousands) 0.40 *** 0.42 ***
(0.01) (0.01)
Bathrooms 0.00 0.01
(0.01) (0.01)
Bedrooms -0.04 *** -0.05 ***
(0.00) (0.00)
Foreclosure -0.23 *** -0.22 ***
(0.01) (0.01)
Constant 11.64 *** 11.80 ***
(0.03) (0.04)
Notes: The total number of observations for the quantile regression
is 15,668. Controls for vintages were included. Fixed effects were
included for each census tract and each year and month houses were
sold. Standard errors, in parentheses, are cluster robust at the
census tract level.
Significance at the 0.10, 0.05. and 0.01 levels is indicated by *,
**, and ***, respectively.
