Difficult to Show Properties and Utility Maximizing Brokers

Authors B r u c e G o r d o n , S e a n P. S a l t e r a n d
Ke n H . J o h n s o n
Abstract This article is the winner of the Real Estate and the Internet
manuscript prize (sponsored by PricewaterhouseCoopers)
presented at the American Real Estate Society Annual Meeting.
Brokers have long believed that difficult to show properties sell
at lower prices and take longer to sell. Where difficult to show
properties are defined as those properties that present
extraordinary difficulties for a broker in arranging or showing
the listing to a particular buyer. Buyers’ recent access to online
real estate applications may make the cost of avoiding these
properties prohibitive to brokers. Employing a hedonic pricing
model and duration modeling techniques, this study finds that
property price and marketing time are not significantly affected
for these properties. The results suggest that brokers possess
limited market power.
Introduction
The majority of properties that are listed and marketed by brokers are not sold by
the property’s listing broker. In most cases, the property in question is sold by a
cooperating broker, either from another firm or another broker within the listing
firm.1 The term ‘‘broker’’, though not technically correct, is used in this study to
represent all the licensed salespeople and associate brokers working within a given
Multiple Listing Service (MLS).
Today, most real estate firms, and their brokers that specialize in the selling of
residential properties, are members of a MLS. Properties marketed by a broker
are placed in the MLS and made available to all other member brokers. The listing
broker makes an array of information about the subject listing available to
potential selling brokers. This information includes, but is not limited to,
particulars such as the property list price, the number of bedrooms, number of
baths, car storage, school zones, exterior and interior amenities, kitchen features
and a listing broker comments.2 Showing instructions are also made available
through the MLS. These instructions typically include owner contact information
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2
1 1 2 G o r d o n , S a l t e r a n d J o h n s o n
and other information, including the presence of a pet, which is designed to
facilitate the showing of the property by potential cooperating brokers.
There is a long held belief among practicing brokers that, ceteris paribus, difficult
to show properties translate into lower prices and longer marketing times. Where
difficult to show properties are defined as those properties that present
extraordinary difficulties for a broker in arranging or showing the listing to a
particular buyer. The idea is simple and incorporates two long accepted economic
paradigms, utility maximization and the law of supply and demand. It is
hypothesized that in order to maximize their utility, showing brokers in effect
create two demand schedules. One schedule reflects the demand for properties that
are not difficult to show. A second schedule reflects the demand for difficult to
show properties.3 The second schedule is located inside the first, creating an
equilibrium price that is lower for difficult to show properties (see Exhibit 1).4 A
corollary to this hypothesis is an extension in property marketing time due to
fewer showings.
Equilibrium Price
Exhibit 1
Price
S
P’
P’’
DND
DD
Q’’ Q’ Quantity
Here DND and DD represent the demand schedules for properties that are not difficult to show and
properties that present difficulties in showing, respectively. The equilibrium price for not difficult to
show properties, P , is greater than the equilibrium price for properties that are difficult to show, P .
D i f f i c u l t t o S h o w P r o p e r t i e s 1 1 3
At one time or another all brokers have been cornered in a bathroom by an
unfriendly dog, chased an escaped cat, found it difficult to arrange a showing time
with an owner who requires an appointment be set in advance, had to drive across
town to pick-up a key for just one of many showings, etc. In addition, brokers
typically arrange several showings and delays at one showing often produce delays
at others. These problems are not trivial and, in fact, produce an increase in the
marginal cost of showing difficult to show properties. Therefore, selling brokers
view the marginal cost of including many of these difficult to show properties
in the feasible set of potential purchases as being greater than the properties’
marginal benefit from inclusion. Accordingly, these properties are dropped
from consideration and thus face a different demand schedule than less difficult
showings.
While this trade-held hypothesis might be true, there is another possible and
competing hypothesis. Specifically, the trade held hypothesis implicitly assumes
that showing brokers exert some significant level of market power over consumers
of real estate services. Where market power is operationally defined as the ability
of brokers to suppress otherwise competitive market forces, specifically many
sellers and buyers of a relatively homogenous product, which should yield a single
equilibrium price. Under this market power scenario, selling brokers can maximize
their own utility by avoiding difficult to show properties at the expense of market
participants resulting in price reductions and extended marketing times for difficult
to show properties. If on the other hand, the market power of these brokers is
limited, property price and marketing time may not be affected.5
A recent spate of articles, including but not limited to Baen and Guttery (1997),
Tuccillo (1997), Bardhan, Jaffee and Kroll (2000), Bond, Seiler, Seiler and Blake
(2000) and Jud, Winkler and Sirmans (2002), and, hint at a possible source for
limiting brokers’ market power. All of these articles address to one extent or
another the impact of the Internet on residential brokerage. Baen and Guttery along
with Tuccillo explicitly argue that the growing use of Internet real estate
applications reduce information and transactions costs, speed transactions and will
ultimately lead to lower commissions by reducing the demand for brokerage
services.
Today, the Internet provides buyers with numerous online real estate applications
that allow buyers to prescreen potential properties via searchable databases and
virtual tours, as well as to prequalify themselves with mortgage calculators.
Showing brokers, fearing that buyers may learn of potential purchases from
Internet listings of other broker listed properties and no longer being able to count
on the inefficient nature of media, such as classified advertising, could now find
it sensible to include difficult to show listings in the feasible set of showings
presented to buyers. Said another way, the rational utility maximizing broker,
seeing that the expected marginal benefit of showing difficult to show properties
is greater than the marginal cost of avoiding these properties, now includes them
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2
1 1 4 G o r d o n , S a l t e r a n d J o h n s o n
in the feasible showing set. In effect, this creates one demand schedule for all
properties and a single uniform equilibrium price no matter what the showing
instructions for a listed property (see Exhibit 2).6 The extended corollary from
this line of thought suggests that marketing time for these difficult to show
properties would not be impacted by their showing instructions.
Are the prices of difficult to show listings affected adversely? Does the marketing
time for these properties extend, relative to similar properties? Alternatively, are
market forces afoot, perhaps online real estate applications, which can change the
showing patterns of rational utility maximizing brokers? These are the questions
examined in this study.
Employing comparable sold data from the Montgomery, Alabama area, this study
investigates the efficacy of the above arguments. A contiguous area is sampled in
order to create a homogenous product and lower overall variability in the housing
product. A hedonic pricing model is developed to investigate the effect, if any, of
difficult to show listings on property selling price. In addition, duration modeling
techniques are used to investigate for ‘‘time on the market’’ effects. Three proxies
Market Forces on Equilibrium Price
Exhibit 2
Price
S
P
DAll Listings
Q Quantity
Market forces exert themselves and consumers view only one demand schedule creating one
equilibrium price.
D i f f i c u l t t o S h o w P r o p e r t i e s 1 1 5
are posited for difficult to show properties in the hedonic pricing model and in
the duration model. PETS, KEYINOFF and APP represent properties that have
pets, require key retrieval from the listing broker and require prearranged showing
times, respectively. Any one, or a combination of these three showing instructions,
increases the difficulty of arranging a showing for the broker. In the trade held
hypothesis, the proxies for difficult to show properties will sign negative and
significant in the pricing model and positive and significant in the duration model.
Conversely, if the competing hypothesis holds, the proxies for difficult to show
properties instituted as controls in the models should prove benign.
Literature Review
The literature is replete with studies that examine the impact of brokers on
residential property price. In the interest of brevity and exposition, this study does
not digress into a detailed discussion of these works. Interested readers can consult
Yavas (1994) and Benjamin, Jud and Sirmans (2000) for a detailed explanation
of this area of research.
The impact of brokers on property time on the market is less extensive. These
studies can be broken into two distinct categories based on the methodology
employed. Most early, and some more recent studies, attempted to model property
marketing time via ordinary least squares (OLS) estimation. For representative
studies employing OLS estimations, the reader should consult Belkin, Hempel and
McLeavy (1976), Janssen and Jobson (1980), Asabere, Huffman and Mehdian
(1993), and Allen, Faircloth, Forgey and Rutherford (2000), among others.
However, the estimation of time on market models is evolving. Duration modeling
techniques are becoming the standard. These techniques, though new and not
completely accepted, recognize certain deficiencies in employing OLS modeling.
For example, OLS modeling is ill suited because of non-normality of the error
term, which can lead to bias in the model’s estimates of coefficients. Kiefer (1988)
provides an excellent presentation on OLS verses duration methodology. For a
representative sample of works that employ duration modeling, the reader can
consult Yang and Yavas (1995), Jud, Seaks and Winkler (1996) and Johnson,
Salter, Zumpano and Anderson (2001), among others.
The literature virtually ignores questions concerning detailed MLS marketing
information. Haag, Rutherford and Thomson (2000) provide one exception to this
rule and investigate the impact of broker comments found in the remarks section
of MLS listing on property price and marketing time. The study finds that
negative, as well as some positive comments, are associated with a price reduction.
The authors suggest these results indicate that many comments offered by brokers,
in the remarks section of MLS listing, are more hype than substantive. To date,
however, there has not been an examination of property showing instructions on
either property price or marketing time. This study addresses this issue.
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2
1 1 6 G o r d o n , S a l t e r a n d J o h n s o n
Data and Methodology
Data
Two data sources provide the essential data for this study. The Montgomery Area
Association of Realtors MLS provides data on housing characteristics, such as the
number of bedrooms, baths, and other physical characteristics including the
variables of interest PETS (properties with pets), KEYINOFF (properties that
require a key to be retrieved to be shown) and APP (properties which require an
appointment time be arranged with the owner prior to showing). The Montgomery
County Tax Assessor Office provides the needed information on property size
(square footage) and age.
The original data set consisted of all conventional residential closings (2,716) that
occurred during the calendar year 1998 in Montgomery, Alabama. In order to
insure that the data contained a complete set of housing characteristics for each
observation, observations that did not appear in both databases were eliminated.
Next, obvious data entry errors from the MLS database such as negative time on
the market, zero bedrooms or baths, no indication of type of siding, etc were also
eliminated. In addition, given that this study is concerned with the impact of
difficult to show properties, newly constructed and vacant properties were
eliminated. Finally, in a further effort not to muddle the question of the market
being studied, any tenant occupied properties were also eliminated.
This results in a final database of 945 observations on which this study is
conducted. Descriptive statistics and a legend for the variable definitions are
presented in Exhibits 3 and 4, respectively.
Hedonic Pricing Model
In order to test for the impact of difficult to show properties on property price,
the following hedonic pricing model is specified.
LnSP LnAGE LnSQFT LnBED
0 1 2 3
LnBATH LEE LANIER CARVER
4 5 6 7
GAR CPT FP GB
8 9 10 11
SEPSHOW POOL DOUBOVN
12 13 14
EIFS PETS KEYINOFF
15 16 17
APP . (1)
18
D i f f i c u l t t o S h o w P r o p e r t i e s 1 1 7
Summary Statistics
Exhibit 3
Variable Mean Median SE Mean
SP 118,329 103,000 1,802
TOM 82.640 64.000 2.260
AGE 21.640 19.000 0.581
SQFT 1,833 1,706 17.800
BED 3.237 3.000 0.020
BATH 2.207 2.000 0.020
JD 0.516 1.000 0.016
LEE 0.335 0.000 0.015
LANIER 0.132 0.000 0.011
CARVER 0.016 0.000 0.004
GAR 0.289 0.000 0.015
CPT 0.231 0.000 0.014
DRIVE 0.480 0.000 0.017
FP 0.782 1.000 0.014
GB 0.307 0.000 0.015
SEPSHOW 0.306 0.000 0.015
POOL 0.133 0.000 0.011
DOUBOVN 0.094 0.000 0.010
EIFS 0.048 0.000 0.007
PETS 0.091 0.000 0.009
KEYINOFF 0.018 0.000 0.004
APP 0.119 0.000 0.011
N 945
N - PETS 86
N - KEYINOFF 17
N - APP 112
In this model, LnSP represents the natural log of the sales price. The regressors
LnAGE, LnSQFT, LnBED and LnBATH act as continuous predictors representing
the property’s age, square footage, number of bedrooms and number of bathrooms,
respectively. Location proxies are estimated through the use of four high school
zones. There are three indicator variables in the model: LEE, LANIER and
CARVER, which are equal to one if the property is located in the Lee, Lanier or
Carver High School zones, respectively. They are zero otherwise. The reference
indicator, which is omitted from the model, is Jefferson Davis High School (JD)
and is reflected in the constant term. The model also includes indicator variables
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2
1 1 8 G o r d o n , S a l t e r a n d J o h n s o n
Legend for Variable Definitions
Exhibit 4
SP Contract sales price of the property;
TOM Time on market in days;
AGE Age of the property;
SQFT Square footage of the property;
BED Number of bedrooms in the property;
BATH Number of baths in the property;
JD One if the property is in the Jefferson Davis school zone, zero otherwise;
LEE One if the property is in the Lee school zone, zero otherwise;
LANIER One if the property is in the Lanier school zone, zero otherwise;
CARVER One if the property is in the Carver school zone, zero otherwise;
GAR One if the property has a garage, zero otherwise;
CPT One if the property has a carport, zero otherwise;
DRIVE One if the property has a driveway only, zero otherwise;
FP One if the property has a fireplace, zero otherwise;
GB One if the property has a garden bath, zero otherwise;
SEPSHOW One if the property has a shower separate from the tub, zero otherwise;
POOL One if the property has a pool, zero otherwise;
DOUBOVN One if the property has a double oven, zero otherwise;
EIFS One if the property is clad in EIFS, zero otherwise;
PETS One if the property has pets, zero otherwise;
KEYINOFF One if retrieving key from listing broker is required, zero otherwise; and
APP One if setting an appointment is required, zero otherwise.
for three types of parking. GAR (garage) and CPT (carport) are specified explicitly
in the model, while DRIVE (driveway only) is specified implicitly in the base
term. GAR, CPT and DRIVE take on values of one if the property has garage,
carport or driveway only parking. Otherwise, the observation in question receives
a value of zero.
In addition to these conventional regressors, controls for varying levels of quality
among the differing properties are placed in the model. This is done through the
inclusion of five quality variables. FP (fireplace), GB (garden bath), SEPSHOW
(separate shower), POOL (in-ground swimming pool) and DOUBOVN (double
oven) are characteristic of higher-quality properties in the sample area and are
available for use in the model. Each of these indicators receives a value of one if
the quality characteristic is present and zero otherwise.
Johnson, Salter, Zumpano and Anderson (2001), employing an expanded version
of this data set, find that EIFS (exterior insulation and finish systems) is positive
D i f f i c u l t t o S h o w P r o p e r t i e s 1 1 9
and significantly related to SP (sales price). EIFS is an exterior siding that has
encountered a significant amount of bad press as of late. The siding, though
originally hailed for its superior insulation quality, has proven to be permeable to
water, which has in turn caused significant structural damage over time.
Seemingly, EIFS-clad properties should be discounted by the market. However,
given the presence of uninformed buyers in the test market and EIFS’s high
correlation with quality, the positive and significant relationship between EIFS and
price is not surprising. Accordingly, EIFS is placed in the model.
Finally, PETS, KEYINOFF and APP, which control for the presence of pets, key
retrieval by the showing broker and arranged appointments, respectively, act as
proxies for difficult to show properties in the pricing model. All else being equal,
any one, or a combination of these three categories of showing instructions,
increases the difficulty in showing for a broker, and lowers the broker’s expected
utility if included in the feasible set of potential purchases due to higher marginal
cost. If the offered trade held hypothesis holds, the coefficients of these predictors
will be negative and significant. If on the other hand, rational utility maximizing
selling brokers now find it beneficial to include difficult to show properties in the
feasible showing set, perhaps because of expanded Internet use, these predictors
will be nonnegative.7
Duration Model
As mentioned earlier, the standard for modeling a property’s time on the market
(TOM) has shifted from OLS modeling to employment of nonlinear techniques.
Often referred to as duration modeling, this methodology provides a measure of
the probability of time on the market for a property. The following Weibull
operational model is specified.
exp(X ) LnAGE 2LnSQFT 3LnBED
0 1
4LnBATH 5LEE 6LANIER
CARVER GAR 9CPT 10FP
7 8
11GB 12SEPSHOW 13POOL
14DOUBOVN 15EFIS 16PETS
KEYINOFF APP . (2)
17 18
The flexible nature of the Weibull allows the function to be either monotonically
increasing or decreasing. If the scale parameter ( ), which indicates duration
dependence, is greater than one, positive duration dependence is indicated. That
is to say, the probability of a property selling is increasing through time. If the
scale parameter is less than one, the chance of a property selling decreases through
time. Interestingly, a unique quality of the Weibull hazard function occurs when
the scale parameter equals one. If equals 1, the specified hazard function reduces
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2
1 2 0 G o r d o n , S a l t e r a n d J o h n s o n
to the exponential hazard function (i.e., the probability of a sale is constant through
time). The interested reader can consult Greene (1997) for a detailed discussion
on this topic, in addition to the works mentioned earlier in the Literature Review
section. All of the independent predictors in this model are as defined in the
hedonic pricing model.
If, as brokers suspect, difficult to showing properties lead to extended marketing
times, the regressors PETS, KEYINOFF and APP will be positive and significant.
Conversely, nonpositive results will suggest a single demand schedule and uniform
TOM no matter what the property showing instructions. Benign results will also
suggest limited market power on the part of brokers.
Empirical Results
Exhibits 5 and 6 formally report the findings of the specified pricing and duration
models. Exhibit 7 reports the results of a nonspecified OLS model of TOM as
Hedonic Pricing Model
Exhibit 5
Predictor Coef Std. Dev. T P VIF
Constant 8.742 0.214 40.880 0.000
LnAGE 0.030 0.010 2.990 0.003 1.4
LnSQFT 0.258 0.029 8.870 0.000 1.3
LnBED 0.351 0.043 8.110 0.000 1.3
LnBATH 0.450 0.033 13.680 0.000 1.5
LEE 0.027 0.017 1.610 0.107 1.2
LANIER 0.086 0.024 3.530 0.000 1.3
CARVER 0.147 0.060 2.450 0.014 1.1
GAR 0.217 0.018 11.730 0.000 1.4
CPT 0.105 0.019 5.630 0.000 1.2
FP 0.133 0.020 6.750 0.000 1.3
GB 0.067 0.019 3.470 0.001 1.5
SEPSHOW 0.135 0.020 6.770 0.000 1.6
POOL 0.090 0.022 4.050 0.000 1.1
DOUBOVN 0.071 0.026 2.710 0.007 1.1
EIFS 0.158 0.036 4.340 0.000 1.2
PETS 0.034 0.026 1.330 0.184 1.1
KEYINOFF 0.023 0.055 0.410 0.683 1.0
APP 0.014 0.023 0.630 0.531 1.1
945; Adj. R2 71.8; and Adj. R2
Notes: The dependent variable is LnSP. N 71.2.
D i f f i c u l t t o S h o w P r o p e r t i e s 1 2 1
Duration / Weibull
Exhibit 6
Predictor Coef Std. Error Z P
Intercept 3.162 0.853 3.710 0.000
LnAGE 0.022 0.040 0.560 0.576
LnSQFT 0.128 0.117 1.090 0.275
LnBED 0.173 0.182 0.950 0.340
LnBATH 0.180 0.127 1.420 0.157
LEE 0.140 0.068 2.060 0.040
LANIER 0.110 0.100 1.100 0.272
CARVER 0.297 0.240 1.240 0.216
GAR 0.060 0.076 0.790 0.428
CPT 0.021 0.076 0.270 0.786
FP 0.124 0.081 1.530 0.127
GB 0.053 0.078 0.680 0.494
SEPSHOW 0.170 0.080 2.140 0.032
POOL 0.018 0.089 0.200 0.842
DOUBOVN 0.103 0.106 0.970 0.331
EIFS 0.279 0.147 1.890 0.058
PETS 0.023 0.106 0.220 0.829
KEYINOFF 0.209 0.228 0.910 0.361
APP 0.059 0.094 0.630 0.528
1.122 0.023
Note: The dependent variable is LnTOM. The 95% CI of the scale parameter is 1.066 1.182
implying the Weibull distribution is preferable to the exponential distribution. N 945. The log
likelihood 1412.2.
well. The explanatory power (R2) of the pricing model is 71.8%, and the model
is highly significant, as indicated by its F-Statistic (not reported) of 130.95
.00011). The coefficients for the control variables, excluding the
(p-value
variables of interest, are all correctly signed and statistically significant. The
model’s variance inflation factors (VIF) are reported formally in Exhibit 5 and are
all within accepted levels of tolerance.
The continuous regressor (AGE) is negative and significantly related to SP as
expected. The remaining continuous regressors SQFT, BED and BATH are all
positive and significantly related to SP, again as expected. As in all hedonic
pricing models, there is a need to control for property location. Specifically, the
Jefferson Davis (JD) school zone is needed to represent the preferred school zone
in the study area. Consultation with local real estate professionals indicates that
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2
1 2 2 G o r d o n , S a l t e r a n d J o h n s o n
Duration / OLS
Exhibit 7
Predictor Coef Std. Dev. T P VIF
Constant 2.524 1.215 2.080 0.038
LnAGE 0.081 0.056 1.440 0.149 1.4
LnSQFT 0.112 0.165 0.680 0.497 1.3
LnBED 0.082 0.246 0.330 0.738 1.3
LnBATH 0.244 0.187 1.310 0.191 1.5
LEE 0.184 0.096 1.910 0.057 1.2
LANIER 0.052 0.139 0.370 0.711 1.3
CARVER 0.629 0.340 1.850 0.065 1.1
GAR 0.185 0.105 1.760 0.079 1.4
CPT 0.023 0.106 0.220 0.828 1.2
FP 0.184 0.112 1.650 0.100 1.3
GB 0.115 0.110 1.050 0.292 1.5
SEPSHOW 0.292 0.113 2.580 0.010 1.6
POOL 0.110 0.126 0.870 0.385 1.1
DOUBOVN 0.006 0.149 0.040 0.966 1.1
EIFS 0.335 0.206 1.620 0.105 1.2
PETS 0.017 0.146 0.110 0.910 1.1
KEYINOFF 0.191 0.315 0.610 0.545 1.0
APP 0.087 0.131 0.670 0.505 1.1
945; R2 3.5; and Adj R2
Note: The dependent Variable is LnTOM. N 1.6.
the general preference ordering of school zones are Jefferson Davis, Lee, Lanier
and Carver, respectively. Consequently, these regressors should sign negative with
increasing orders of magnitude for the zones CARVER, LANIER and LEE,
respectively. These are the results. In addition, the coefficients for the location
proxies are all significantly related to SP.
Driveway only (DRIVE) is specified as the base case for parking type. Properties
with driveway only parking should be the least preferable, while garage parking
should be the most preferable. Thus, the coefficients for GAR and CPT should be
positive and significantly related to SP, with GAR having the greater magnitude.
Again, these results are found.
Turning to the model’s dichotomous controls, the regressors for the presence
of a fireplace (FP), garden bath (GB), separate shower (SEPSHOW), in-ground
swimming pool (POOL), double oven (DOUBOVN) and exterior insulation and
finish systems (EIFS) are all positive and significantly related to SP as expected.
D i f f i c u l t t o S h o w P r o p e r t i e s 1 2 3
Finally, this study examines PETS, KEYINOFF and APP for their impact on
property price. Interestingly, none of these variables are significant. These results
tends to discredit the popular belief among brokers of lower prices for difficult to
show properties and lend support to the competing hypothesis. These results also
have practical implications. Specifically, when confronted with a seller who does
not want to board their pets or insists on being difficult about their property’s
showing arrangements, listing brokers no longer need to worry about lower sales
prices and hence commissions. Selling brokers, on the other hand, would be well
advised to market all properties equally, irrespective of their perceived difficulties
in showing.
This study’s duration model is examined next. A 95% confidence interval on the
scale parameter, 1.066 1.182, suggests positive duration dependence and
that the Weibull model is preferred to the exponential model. The model indicates
three statistically significant factors affecting selling time. Properties located in
the Lee school zone sell significantly faster than properties in other location
proxies. This result is somewhat surprising, given the speculation by brokers in
this market that the JD location proxy is preferable. Interestingly, when combining
this result with the pricing model, this study finds, that while properties located
in the JD school zone receive a pricing premium, they take longer to sell on
average than properties located in the Lee school zone. These combined results
may be indicative of a shift in demand patterns and should prove helpful to local
brokers in the Montgomery area. The control for SEPSHOW is also significant
and negative. This result is expected, as a separate shower is a preference item
among buyers in the study area leading to a shorter selling time.
EIFS on the other hand, is statistically significant and positive, suggesting that
properties with EIFS take longer to market. This result, when combined with the
pricing model, is consistent with the hypothesis put forth in Johnson, Salter,
Zumpano and Anderson (2001) and is not surprising. Specifically, sellers in this
market, perhaps on the advice of their brokers, perceive of the existence of
uninformed EIFS buyers. Therefore, sellers of EIFS-clad properties do not
discount their price and wait for an uninformed buyer. The end result is a
nonnegative pricing impact from the presence of EIFS, but property marketing
time is extended. These results suggest that while sellers of EIFS-clad properties
may not receive a discount in their prices; the present value of their final proceeds
is less due to extended marketing time.8
All three of the proxies for difficult to showing properties are insignificant. These
results suggest that the long held belief of extended marketing times for difficult
to show properties is not warranted.9 Again, these findings lend support to the
competing hypothesis, suggesting limited marketing power on the part of brokers.
Conclusion
This study seeks to examine two competing hypotheses. Under both hypotheses,
selling brokers use a marginal cost benefit analysis to determine which properties
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2
1 2 4 G o r d o n , S a l t e r a n d J o h n s o n
to include in the set of feasible purchases presented to buyers. The first hypothesis
argues that brokers, without scientific investigation, believe that difficult to show
properties receive lower prices while witnessing extended marketing times due to
the increase in the marginal cost of presenting difficult to show properties to
buyers. Two different demand schedules are then presented with the demand
schedule for difficult to show properties being inside the demand schedule for
properties that do not present difficulties in showing. Ceteris paribus, the resulting
equilibrium price for difficult to show properties is lower than the price for other
properties. A corollary to this hypothesis is that fewer buyers will visit these
properties thus extending their marketing time. Significant market power on the
part of brokers is a necessary condition for this hypothesis to hold.
The competing hypothesis suggests that the casual beliefs of brokers may not
hold. Specifically, it may be possible that, due to online real estate applications
such as buyer searchable databases of available properties, virtual tours and
mortgage calculators, brokers may not be able to effectively maximize their own
utility by avoiding difficult to show properties. Under this scenario, a single
demand schedule is present resulting in uniform pricing. A corollary to this
competing hypothesis is that difficult to show properties should not witness
extended marketing times. This would be true if brokers possess limited power
over the market.
Using comparable sold data from the Montgomery, Alabama area, three categories
of difficult to show properties are classified. PETS, KEYINOFF and APP represent
properties that have pets, require key retrieval from the listing broker, and require
prearranged showing times, respectively. Any one, or a combination of these three
categories, increases the difficulty of a sale for the broker. These proxies for
difficult to show properties are specified in a hedonic pricing and duration model
to test for their impact on property price and marketing time.
The statistical results indicate that none of the proxies for difficult to show
properties influence either property price or marketing time. These results suggest
that brokers have limited market power, at least less than previously suspected,
and the casual beliefs of brokers concerning property price and marketing time
for difficult to show properties is misplaced. Buyers are now informationally
empowered as never before, and the logical cause for this reduction in market
power of brokers appears to be the Internet. In addition, the results have practical
applications. Past perceptions of difficult sellers causing longer marketing times,
lower sales prices and thus lower commissions for listing brokers, do not seem to
hold. Selling brokers, on the other hand, would be well advised to show all
properties irrespective of any showing difficulties involved.
A few words of caution seem warranted. First, it would be beneficial to test if the
casual beliefs of brokers ever held. Specifically, did difficult to show properties
experience a price discount and extended time on the market prior to recent Web
technologies? Second, it may be possible that the ratio of difficult to show
properties to properties that do not present difficulties in showing could alter the
D i f f i c u l t t o S h o w P r o p e r t i e s 1 2 5
results of this study. Said another way, the marginal analysis performed by brokers
to determine the feasible set of alternatives presented to buyers could vary
depending on the makeup of available properties. Third, a comparison of the
number of showings across difficult and alternative listings could provide
additional insight. Unfortunately, the data needed for these additional tests is not
available. All comparable data prior to 1998 has been purged from the
Montgomery area MLS system for data storage reasons, making the first two
suggestions impractical. A count of the number of showings for the differing
properties would be possible if all of the listings in the test area were equipped
with electronic lockboxes. This recent technology is becoming available in some
markets but was not available in 1998 for the test market. Taking all of these
limitations into account, the findings in this study, while suggestive of limited
market power on the part of brokers, need additional research to be confirmed.
Notwithstanding these concerns, the empirical results of this study strongly
indicate that difficult to show properties do not experience either a discount in
price or extended marketing time.
Endnotes
1
The data employed for this study contains 945 comparable sales. The listing and selling
broker were the same individual in only 78 cases or approximately 8.3% of the total
number of observations. Consultations with brokers in other markets revealed that, at
least casually, this rate of ‘‘double dipping,’’ as it is often referred to in the trade, seems
reasonable. Quite simply, listing brokers, though possessing certain synergistic
advantages, are overwhelmed by the vast number of competing cooperating brokers in
the market. Finally, the term listing broker, when used in this study, represents the broker
who has listed the property. The terms selling, showing or cooperating broker represents
a broker aiding buyers in locating property.
2
In fact, it is not uncommon for an MLS to have upwards of 300 profiled features available
for any listing. A listing of all of these features would prove exhaustive and in the interest
of brevity and space is omitted.
3
This second demand schedule can be thought of as a tax that has been placed on the
seller of property. The resulting tax incidence debate is another research question worthy
of interest. The dynamics of this question are not so straightforward, however, because
listing and selling brokers interest are not so perfectly aligned. Therefore, this additional
issue is set aside for the moment and addressed in future research in order to concentrate
on the question at hand. Said another way, this work makes the simplifying assumption
of no conflict of interest between the listing and selling brokers.
4
An alternative explanation for this trade held hypothesis is that the supply of properties
is perfectly inelastic in the short run. Exhibit 1 obviously takes a long run view of the
market. The results, however, are robust no matter the elasticity of the supply schedule.
Specifically, P’’ is lower than P’ regardless of the time frame considered for the model.
5
Frew (1987) provides an alternative explanation. Frew argues that brokers, seeking to
maximize their income, may ‘‘holdback’’ prime listings. This strategic move on the part
of listing brokers could easily influence the resulting price and time on market
estimations. However, ‘‘holdbacks’’ are inextricably intertwined with the tax incidence
question outlined in Endnote 3 and are reserved for future research.
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2
1 2 6 G o r d o n , S a l t e r a n d J o h n s o n
6
A point of clarification is perhaps warranted. Some may question the demand side impact
in both Exhibits 1 and 2 as opposed to a supply side effect. This is so because the focus
here is on the rational behavior of brokers who work with buyers and hence impact
demand. Listing brokers impact the market via the supply schedule. Their rational utility
maximizing behavior is held in check in order to investigate the demand side effects in
isolation. See Endnotes 3 and 5 for further explanation.
7
With thanks to an anonymous reviewer, it may be the case that high probability of sale
properties are significantly and positively related to difficult to show properties suggesting
some degree of selection bias. This study, however, explicitly holds all other factors,
including a property’s inherent probability of sale, which might influence a listing’s
marginal cost equal. However, future research concerning difficult to show properties
should investigate this potential bias.
8
Johnson, Salter, Zumpano and Anderson (2001) find EIFS-clad properties on average
experience extended marketing times of an additional 28 days. The average cost of
replacing EIFS siding and any necessary structural damage was not available for the
study. However, conversations with local brokers, home inspectors and contractors
indicated that these repairs far exceeded any lost value in proceeds due to extended
marketing time.
9
Alternate versions of both pricing and duration models were specified, and the results
remained consistent across all versions. Difficult to show properties did not suffer a
pricing discount, and they did not experience extended marketing times.
References
Allen, M. T., S. Faircloth, F. Forgey and R. C. Rutherford, Salesperson Compensation and
Performance in the Housing Market, Working Paper, 2000.
Asabere, P. K., F. E. Huffman and S. Mehdian, Mispricing and Optimal Time on the Market,
Journal of Real Estate Research, 1993, 8:1, 149–56.
Baen, J. S. and R. S. Guttery, The Coming Downsizing of Real Estate: Implication of
Technology, Journal of Real Estate Portfolio Management, 1997, 3, 1–18.
Bardhan, A., D. Jaffee and C. Kroll, A Research Report: The Internet, E-Commerce and
the Real Estate Industry, Fisher Center for Real Estate and Urban Economics, Hass School
of Business, University of California Berkley, 2000.
Belkin, J., D. J. Hempel and D. W. McLeavy, An Empirical Study of Time on Market
Using Multidimensional Segmentation of Housing Markets, Journal of the American Real
Estate and Urban Economics Association, 1976, 4:1, 57–75.
Benjamin, J. D., G. D. Jud and G. S. Sirmans, Real Estate Brokerage and the Housing
Market: An Annotated Bibliography, Journal of Real Estate Research, 2000, 20:1/2, 217–
78.
Bond, M. T., M. J. Seiler, V. L. Seiler and B. Blake, Uses of Websites for Effective Real
Estate Marketing, Journal of Real Estate Portfolio Management, 2000, 6: 203–10.
Frew, J. K., Multiple Listing Service Participation in the Real Estate Brokerage Industry:
Cooperation or Competition, Journal of Urban Economics, 1987, 21:3, 272–86.
Greene, W. H., Econometric Analysis, Third edition, Upper Saddle River, NJ; Prentice-
Hall, 1997.
Hagg, J. T., R. C. Rutherford and T. A. Thomson, Real Estate Agent Remarks: Help or
Hype, Journal of Real Estate Research, 2000, 20:1/2, 205–215.
D i f f i c u l t t o S h o w P r o p e r t i e s 1 2 7
Janssen, C. T. L. and J. D. Jobson, On the Choice of Realtor, Decision Science, 1980, 11,
299–311.
Johnson, K. H., S. P. Salter, L. V. Zumpano and R. I. Anderson, Exterior Insulation and
Finish Systems: The Effect on Residential Housing Prices and Marketing Time, Journal of
Real Estate Research, 2001, 22:3, 289–311.
Jud, G. D., T. G. Seaks and D. T. Winkler, Time on the Market: The Impact of Residential
Brokerage, Journal of Real Estate Research, 1996, 12:3, 447–58.
Jud, G. D., D. T. Winkler and G. S. Sirmans, The Impact of Information Technology on
Real Estate Licensee Income, Journal of Real Estate Practice and Education, 2002, 5:1,
forthcoming.
Kiefer, N. M., Economic Duration Data and Hazard Functions, Journal of Economic
Literature, 1988, 26:2, 646–79.
Tuccillo, J. A., Technology and the Housing Markets, Business Economics, 1997, 32, 17–
20.
Yang, S. X. and A. Yavas, Bigger is not Better: Brokerage and Time on the Market, Journal
of Real Estate Research, 1995, 10:1, 23–33.
Yavas, A., Economics of Brokerage: An Overview, Journal of Real Estate Literature, 1994,
2:2, 169–195.
The authors would like to thank Jim Frew, Don Jud, James Larsen and Daniel Winkler,
among others, for their insightful comments and suggestions provided at the 2001
ARES conference. In addition, we extend our thanks to the Alabama Real Estate
Research and Education Center for its continued support.
Bruce Gordon, University of North Alabama, Florence, AL 35632 or bgordon@
unanov.una.edu.
Sean P. Salter, University of Southern Mississippi, Hattiesburg, MS 39406 or
spsalter@earthlink.net.
Ken H. Johnson, Florida Atlantic University, Boca Raton, FL 33431 or johnson3@
fau.edu.
J R E R Vo l . 2 3 N o s . 1 / 2 – 2 0 0 2