of 16

Auction

Published on April 2017 | Categories: Documents | Downloads: 1 | Comments: 0
22 views

Comments

Content


http://usj.sagepub.com/
Urban Studies
http://usj.sagepub.com/content/42/12/2245
The online version of this article can be found at:
 
DOI: 10.1080/00420980500332007
2005 42: 2245 Urban Stud
Carl R. Gwin, Seow-Eng Ong and Andrew C. Spieler
Auctions and Land Values: An Experimental Analysis
 
 
Published by:
http://www.sagepublications.com
On behalf of:
 
Urban Studies Journal Foundation
can be found at: Urban Studies Additional services and information for
 
 
 
 
http://usj.sagepub.com/cgi/alerts Email Alerts:
 
http://usj.sagepub.com/subscriptions Subscriptions:
http://www.sagepub.com/journalsReprints.nav Reprints:
 
http://www.sagepub.com/journalsPermissions.nav Permissions:
 
http://usj.sagepub.com/content/42/12/2245.refs.html Citations:
 
What is This?
 
- Nov 1, 2005 Version of Record >>
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
Auctions and Land Values: An Experimental
Analysis
Carl R. Gwin, Seow-Eng Ong and Andrew C. Spieler
(Paper first received, October 2003; in final form, May 2005)
Summary. Although urban land development is a well-researched area, there has been
inadequate research on the allocation mechanism for land. This paper offers a new perspective
on the effects of the auction mechanism on land values. By appealing to an experimental setting
to examine the bidding behaviour of developers in repeated first-price sealed bid (FPSB)
auctions where resolution of price uncertainty (development revenue) is deferred, it is shown
that overbidding is pervasive. Participants consistently shave expected profit margins in order to
secure land for development. The actual profit margins reported in all experiments are
significantly lower than the suggested benchmark made known before the experiments.
Consequently, participants tend to experience smaller ex post profits relative to ex ante
expectations. However, we find that repeated bidding does mitigate overbidding to some extent,
as evidenced by lower differences in the winning bid to the second-highest bid. Convergence in
the range of bids occurs over time, but only significantly so when public information in terms of
an expectation of the future property price, is provided. This study offers interesting policy
implications for urban land development.
1. Introduction
The study of urban land development has been
important to studies of the urban form. While
there is an extensive literature on land
reforms, land development processes and
land valuation systems, a fundamental ques-
tion that has yet to receive adequate coverage
in the literature is the allocation mechanism
by which land is released for development.
This paper concerns itself primarily with one
mechanism among many—auctions.
Land auctions are the primary means of
supply in many countries and cities. An import-
ant feature of urban development in Singapore
and Hong Kong, for instance, is the auction of
land for development and investment (Fu and
Ching, 2003). China witnessed its first land
use auctions in Shenzhen in 1987 (Tang, 1989;
Wu and Yeh, 1997) and Shanghai in 1988
(Xie et al., 2002). Land auctions in many
countries are closely watched as a barometer
of the real estate market with regard to market
interest, future price expectations and general
outlook. The modern landscape of land
development in China is shaped by the 1990
Ordinance of Conveyance and Transfer of
Land-use Rights in State-owned Urban Land
that allows transfer of land rights by auction
and tender (Xie et al., 2002).
In an economic framework, auction prices
are the ex post realisation of the ex ante
distribution of private values (Milgrom,
2004; Quan, 1994). Given that land auctions
Urban Studies, Vol. 42, No. 12, 2245–2259, November 2005
Carl R. Gwin is in the Department of Economics, Baylor University, Waco, TX 76798-8003, USA. Fax: 254 710 6142. E-mail: carl_g-
[email protected] Seow-Eng Ong is in the Department of Real Estate, National University of Singapore, Singapore 117566. Fax: 65
6774 8684. E-mail: [email protected] Andrew Spieler is in the Department of Finance, Hofstra University, Hempstead, NY 11549-
1000, USA. Fax: 516 463 4834. E-mail: [email protected] The authors acknowledge funding by the National
University of Singapore Research Project 3972043.
0042-0980 Print=1360-063X Online=05=122245–15 # 2005 The Editors of Urban Studies
DOI: 10.1080=00420980500332007
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
in many countries are held frequently,
pertinent questions for research include the
following. Does bidding through auctions
result in higher land values? Do bidders
learn from past experience? Should policy-
makers provide relevant public information
to influence bidding behaviour or should the
government maintain a laissez faire attitude?
This paper seeks to examine these import-
ant questions within the context of a close
tender or first-price sealed bid (FPSB) mech-
anism by appealing to an experimental
approach.
1
Earlier works on single-object
auctions (Kagel and Levin, 1986, 1991;
Hansen and Lott, 1991; Lind and Plott,
1991) suggests that FPSB auctions lead to
the winner’s curse. The winner’s curse
applies to common value auctions while land
auctions are often viewed as private value
auctions (Ooi et al., 2005). A logical infer-
ence, however, is that bidders for land over-
pay and, as a result, land values are likely to
be inflated when the auction mechanism is
used relative to the case where the auction
is not used as an allocation mechanism.
However, what is not clear is whether and to
what extent overbidding prevails under a
repeated FPSB land auction mechanism when
bidders learn from past experience. The
auction of land and real estate is relatively
unexplored, with the exceptions of Colwell
and Yavas (1994) andYavas et al. (2000, 2001).
This study examines how FPSB auctions
affect the formation of land prices over time
relative to property prices to answer the fol-
lowing research questions. First, overbidding
in a land auction depends not only on the
experience of the bidders (Samuelson and
Bazerman, 1985; Kagel and Levin, 1986)
but, perhaps more importantly, also on the
expected profit margin. In bidding for land,
developers incorporate their expected profit
margin under which they are compensated
for bearing risk given uncertain property
prices.
2
Does competition in the FPSB land
auction mechanism lead developers to
reduce their expected profit margin and, in
so doing, overbid for land?
Secondly, the on-going nature of land
auctions means that auctions are carried out
periodically (quarterly or semi-annually).
Given that property development is typically
a rather protracted process (spanning at least
1–2 years), how will deferred revenue realis-
ation and repeated auctions affect a develo-
per’s capacity to bid? Do developers learn
from experience and make less aggressive
bids eventually?
3
If so, how would this
affect the formation of land prices over time?
Finally, there is an issue pertaining to
whether information, defined as expected
property price, should be made public. Can
the relevant authorities reduce the overbid-
ding by making information pertaining to
expected property prices available to the
public?
In the absence of empirical evidence on
overbidding and the formation of land prices
under alternative auction mechanisms, we
appeal to experimental economics as the
natural and arguably the only framework for
rigorous analysis and policy evaluation of
land auctions. Our experiment shows that the
bidders consistently overbid in auctions.
Participants consistently shave expected
profit margins in order to secure land for
development. The actual profit margins
reported in our experiments are significantly
lower than the suggested benchmark made
known before the experiments. Consequently,
participants tend to experience smaller ex post
profits relative to ex ante expectations.
However, we find that repeated bidding does
mitigate overbidding to some extent as the
difference between the winning bid and the
second-highest bid declines over time.
Convergence in the range of bids occurs
over time, but only significantly so in scen-
arios where information, either in terms of
the future property price or future supply, is
provided to the public. This experimental
result provides evidence that supports the
idea that publicising information in the
form of the lowest private signal mitigates
excessive bidding.
The remainder of the paper proceeds as
follows. Section 2 briefly reviews the
literature on auctions. Section 3 describes
the experimental research methodology.
Hypotheses are laid out in section 4 and
2246 CARL R. GWIN ET AL.
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
results are discussed in the section 5. Section 6
concludes.
2. Brief Literature Review
Overbidding occurs from judgement failure
when actual bids exceed theoretical predic-
tions in common value auctions where the
auctioned item is worth roughly the same to
each bidder, but where no bidder knows
the true common value prior to making
their bids. Bidders win only if they have the
highest private signal of the land value and
the highest estimate is likely to be an overes-
timate of the value. The theoretical prediction
is the Nash equilibrium solution under risk
neutrality (RNNE) developed by Wilson
(1977) and Milgrom and Weber (1982). The
deviation from the RNNE is overbidding or
the winner’s curse.
Kagel and Levin (1986) show that public
information can increase average revenue
from the supplier’s view-point (increase
overbidding) if the number of bidders is
small, but can decrease average revenue
(reduce overbidding) if the number of bidders
is large. Goeree and Offerman (2002)
examine efficiency issues between private
and common value auctions in an experimental
setting. Milgrom (2004) provides an excellent
review of theoretical work on auctions as
well as practical applications. However, little
theoretical and empirical work exists on the
effect of repeated auctions on bidder behaviour
and price formation.
Kagel and Richard’s (2001) work is closest
to the current study in that they consider
whether experienced bidders can overcome
the winner’s curse in FPSB common value
auctions. Kagel and Richard find that experi-
enced bidders are more successful at avoiding
the winner’s curse, but that they still do not
earn as much as they should under the
RNNE. This paper proceeds in a fashion
similar to that of Kagel and Richard in that it
adopts the same theoretical foundation—
from Wilson (1977)—and employs similar
bid functions to study how bidders behave in
repeated auctions. While Kagel and Richard
study generic auctions, our focus is on the
distinctive institutional characteristics of land
auctions. Our results allow us to analyse
policy implications that are specific to land
development.
Similarly, research on land auctions is
limited. A notable exception is Colwell and
Yavas (1994) who examine agricultural land
auctions that are often used to convert land
to urban uses. Most of the existing work
focuses on private value auctions for houses
(Adams et al., 1992; Vanderporten, 1992;
Lusht, 1996; Mayer, 1995, 1998; Dotzour
et al., 1998; Allen and Swisher, 2000). One
interesting study is Lusht (1996), in which it
is established that English auctions for mid-
to high-priced houses in the US extracted
higher prices than private negotiations. Quan
(1994) provides a thorough review of the
theory and practice of real estate auctions.
3. Research Methodology
Experimental economics provides the ideal
platform to examine issues on bidding
behaviour. The advantages of experimental
economics are well established. First, exper-
iments can be useful for policy implications
and decision-making. Experiments have
been conducted to generate data that might
influence a specific decision (Plott, 1982,
1987). Secondly, experiments allow research-
ers to discover empirical regularities in areas
for which existing theory has yet to be fully
developed (McCabe et al., 1993; Friedman,
1993). Thirdly, experiments provide an ideal
platform to test competing theories and pre-
dictions (Fiorina and Plott, 1978). Experi-
mental economics is now an established area
of study in mainstream economics (Palfrey
and Porter, 1991).
An experimental setting is the best available
framework to study issues with land auctions
and to assess policy implications when there
is no solid empirical evidence pertaining to
the effects of capital constraints, supply uncer-
tainty and the value of public information. The
experimental framework for this research is
largely simplified. For instance, we assume
that developers are atomistic in that they have
no individual influence over property prices.
AUCTIONS AND LAND VALUES 2247
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
Moreover, we assume constant development
costs in this experiment. In addition, we
recognise that the development process is
complex with many institutional and infor-
mational constraints. While these assumptions
are likely to be violated in practice, we make
these assumptions to facilitate analysis and to
obtain base case results for future comparisons
when these assumptions are relaxed.
Bidders in the experiments do not face any
capital constraints and supply uncertainty. We
assume that the supply of land is exogenous to
the bidding process and does not react to
property prices. To test how providing infor-
mation to the public affects bidding behaviour
and price formation, bidders (participants)
will be divided into two groups—one group
that operates under no public information
(scenario A) and a second group that receives
public information (scenario B). The public
information will be in the form of a signal of
expected property price.
The emphasis of this study is on analysing
the bidding behaviour of participants and the
formation of land price over time where
bidding occurs at regular intervals. There are
8 participants in each of the 2 groups for a
total of 16 participants for each scenario.
4
Participants comprise undergraduate students
at the National University of Singapore.
While the use of students may be a cause for
concern, Dyer et al. (1989) contrasted the
behaviour of professionals or experts with
naı ¨ve students and found that the experts
were as prone to the winner’s curse as the
naı ¨ve subjects.
5
The number of participants
reflects the average number of developers
who bid at land auctions and also allows for
possible attrition arising from bankruptcies.
6
Studies have shown that the gains in efficiency
from increasing the number of participants
decline sharply as the number of participants
is increased (Gresik and Satterthwaite, 1986;
Satterthwaite and Williams, 1993). These
studies advocate the use of the smallest
number of participants necessary to carry out
the experiment.
The auction mechanism will be the first-
price sealed bid system and all participants
will be endowed with an equal amount of
initial capital. In the first and third scenarios,
participants have free access to additional
funds. Property prices are generated by
random processes with time-trends. A fixed
quantity of land (1 lot) is made available for
auction each period. Each lot of land can be
developed into a fixed number of dwelling
units. The development cost per land lot will
be constant.
7
The development project will
take four periods to complete and revenue
from the sale of the completed property units
will depend on the prevailing property prices
at the end of four periods.
8
At the commence-
ment of each round of bidding, the current
property price and a history of past prices
will be made public. All bidders will also
receive an individual private signal of the
property price at the end of four periods.
These signals are generated randomly and
differ across bidders to simulate imperfect
information or perspective of future property
prices.
9
Bidders decide on the profit margin
they expect and submit sealed bids.
Bids will be revealed to all bidders after all
bids are submitted,
10
and the highest bid wins.
The experiment then moves on to the next
round of bidding. During the fifth round, the
winner of the auction in the first round will
realise his/her revenue and profits or losses
based on the prevailing property price. Partici-
pants who do not win any bids earn zero return
on their capital. There is no prior announce-
ment of the final round to avoid excessively
aggressive bidding behaviour.
Cumulative profits will be computed at the
end of each round,
11
and cash compensation
will be calculated based on a pro rata basis.
A critical issue in experimental economics is
experimental control to ensure that results
are reliable and reproducible (Jamal and
Sunder, 1991; Smith and Walker, 1992).
Cash is the most frequently used reward
(Friedman and Sunder, 1994)
12
and is found
to be highly effective in providing experi-
mental control over participants’ induced
characteristics.
13
Participants who overbid can experience
losses. If the cumulative losses exceed a pre-
specified amount, the participant is deemed
bankrupt and will not be allowed to continue
2248 CARL R. GWIN ET AL.
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
bidding and forfeits all rewards. This is
necessary to maintain realism and to simulate
actual market conditions (Friedman and
Sunder, 1994). To control for risk aversion,
we appeal to the use of binary lottery to
infer subjects’ risk tolerances (Roth and
Malouf, 1979). Potential participants take a
simple lottery quiz
14
and participants who
are risk-neutral are chosen for the game.
The experiment is carried out in two stages.
The first stage entails a pilot experiment with
limited rounds of bidding. The pilot is con-
ducted for both scenarios. The pilot exper-
iment allows us to modify, refine and test
our parameter values. The second stage is
the actual experiment conducted separately
on a different set of participants. Each
session lasts no more than 4 hours with two
breaks in between. All results reported are
from the actual experiment conducted.
4. Hypotheses
The central hypothesis is that winning bids
tend to reduce profit margins, a manifestation
of overbidding behaviour. Stated differently,
the main hypothesis predicts that the land
price (winning bids) will be higher than the
expected price under Risk Neutral Nash
Equilibrium and benchmark profit margin
(henceforth referred to as the RNNE land
price).
15
The RNNE is computed as follows.
Suppose N is the number of bidders and e is
the idiosyncratic error term. In other words,
the private signal s
i
is distributed uniformly
over the true value V
Ã
½V
Ã
À e, V
Ã
þ eŠ.
Further, the lowest private signal revealed
(public information) is

s. In this case, the
RNNE bid for the ith bidder (b
i
) will be
b
i
¼ s
i
À e þ y (1)
where, y is given by
y ¼
2e
N þ 1
exp À
N
2e
½s
i
À (s þ e)Š

(2)
We propose to test the overbidding hypothesis
through a series of hypotheses.
H1: Competition between developers leads to
overbidding.
Overbidding can be measured in many ways.
First, it can be measured in terms of the
expected profit margin incorporated by the
developers at the time of the bidding.
Secondly, overbidding can also be measured
in terms of the premium of the winning bid
over the second-highest bid. Finally, overbid-
ding can be reflected in the difference between
actual and hypothetical (based on RNNE land
price where applicable) realised profits. The
realised profit reflects more than just over-
bidding because it depends on the deviation
of the actual property price from the bidder’s
expectation thereof.
H2: Public information reduces overbidding
and reduces the deviation from expected
prices under RNNE.
In scenario B, public information is the lowest
private signal or the lowest future property
price.
16
H3: Convergence in bids occurs over time as
participants learn from their bidding
experience.
To test this hypothesis, we regress the range of
bids against time and estimate the coefficient
on time. Actual bids will be evaluated with
reference to the theoretical or Risk Neutral
Nash Equilibrium (RNNE) land price, com-
puted assuming a 30 per cent profit margin.
5. Results and Analysis
In this section, we first define the terms used in
the presentation of our results. Specifically, we
formally define the winning bid premium,
average profit margin and realised profit. We
then address each hypothesis in turn.
The winning bid premium in period t
(PREM
t
) is defined as the difference between
the winning bid and the second-highest bid
divided by the second-highest bid.
PREM
t
¼
highest bid
t
À second À highest bid
t
second À highest bid
t
(3)
The average profit margin for the ith partici-
pant in period t (APM
it
) is the expected
property value as given in the private signal
AUCTIONS AND LAND VALUES 2249
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
less the bid and the development cost
expressed as a percentage of the cost (constant
over time and across bidders). So
APM
it
¼
S
it
À B
it
À C
C
(4)
where, B
it
is the bid submitted and s
it
is the
private signal received by the ith participant
in period t. If the public information is the
lowest private signal, then the average profit
margin is computed using the public signal
plus the error term (e) less the bid and devel-
opment cost. The simplification is preferred
over using the RNNE bid since we cannot be
sure that participants are really risk-neutral
even after controlling for risk aversion.
17
Moreover, the APM estimate of the true
future price is straightforward to compute.
The realised profit for the ith participant
arising from the bid in period t (RP
it
) is the
actual property price in period t þ 4 (P
tþ4
)
less the bid at t and development cost,
expressed as a percentage of the cost
RP
it
¼
P
tþ4
À B
it
À C
C
(5)
Overbidding in Repeated Bidding (H1)
Does overbidding persist under repeated
rounds of bidding? Table 1 shows the
average profit margin, premium and realised
profit under scenarios A and B and the
RNNE benchmark. The first point to note is
that the actual land prices (winning bids) gen-
erated from the experiment are higher than the
theoretical land prices. The difference in
means test clearly shows that the land prices
are significantly different from the theoretical
prices.
The average profit margins (expressed as a
percentage of the prevailing property price)
for both scenarios are much lower than the
suggested 30 per cent. The means are 8.44
per cent and 16.31 per cent for scenarios A
and B, both of which are significantly lower
than 30 per cent based on the t-statistics
shown in the last column of Table 1. Hence,
the null in the first hypothesis is rejected in
that the average profit margin is significantly
lower than 30 per cent (see Figure 1). The
same conclusion is made for the other scen-
arios in that the average profit margins are,
in all instances, below the 30 per cent
benchmark.
18
The effect of aggressive bidding behaviour
is manifested in the higher land prices. Over
42 periods in scenario A, the land price
(winning bid) ranged from $48.95 to $117.70
compared with the theoretical land price
range of $31.73–73.29 (see Figure 2). The t-
statistic from a difference in means test
between the actual and theoretical land prices
is 244.13. The means are statistically different
at the 1 per cent significance level. Similarly,
the actual land price under scenario B is
significantly higher than the theoretical
values. The first effect of aggressive bidding
is the tendency to reduce the expected profit
margin and raise land prices.
The second effect is that the realised profits
of developers are reduced compared with the
benchmark case. The mean realised profit
(expressed as a percentage of cost) under
scenarios A and B are 6.712 per cent and
2.004 per cent, respectively. These are low
compared with the theoretical benchmark
mean realised profits of 73.59 per cent and
86.37 per cent (see Figure 3). Although the
differences in means are not statistically
different, the experiment shows that actual
losses are incurred while no losses would
have been incurred if developers use the 30
per cent profit margin as a guide.
To examine further this feature, we
compute the number of times the winning
bid resulted in negative realised profits. In
scenario A, 12 winning bids out of 42 (28.57
per cent) resulted in negative realised profits.
The corresponding figure for scenario B is
17 out of 40 (42.5 per cent).
The premium over the second-highest bid
ranged from a minimum of 0.01 per cent to a
maximum of 13.78 per cent while averaging
3.15 per cent and 5.48 per cent for scenarios A
and B respectively. The theoretical premiums
averaged 3.72 per cent and 1.45 per cent respect-
ively (see Figure 4). The results from the differ-
ence in means test are unable to reject the null
hypothesis for scenario A (t ¼ 0.01). However,
the mean premium in the actual experiment for
2250 CARL R. GWIN ET AL.
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
scenario B(t ¼ 3.73) is significantly higher than
the theoretical mean premium. We observe that
the premium tends to decline over time and
interpret this as evidence that experience and
repeated bidding tend to mitigate overbidding
with.
An interesting analysis is to examine the
minimum expected profit margin in the
cross-section of developers. The minimum
expected profit margin is backed out from
the private signal given to each individual
developer. It is an indicator of the aggressive-
ness of the developer as well as the reliance
that the developer places on the private
signal.
19
In scenario A, there is no public
information and the developers have only
their private signals. Figure 5 shows the
minimum profit margin for scenario
A. While the developer with the minimum
profit margin is not necessarily the developer
with the highest bid,
20
the graph shows the
extent to which the participants are willing
to shave their profit margin in order to win a
bid.
The flip side of the reliance question is to
examine how often participants with the
highest private signal actually win the bid.
In scenario A, participants with the highest
private signal win 17 times out of 42 (40
per cent). The hit ratios (defined as the
number of times the highest theoretical
bidder actually wins the bid divided by the
total number of rounds) are shown in the
Appendix.
Table 1. Winning bid, average profit margin, premium and realised profit for scenarios A and B:
experimental versus theoretical results
Experiment Theoretical
T-statistic for difference in
means
Scenario Scenario A Scenario B Scenario A Scenario B Scenario A Scenario B
Periods 42 40 42 40
Winning bid (land price)
Average 87.11 84.68 53.71 47.43 244.13
ÃÃ
237.65
ÃÃ
Median 95.05 91.79 57.05 51.69
Standard deviation 22.36 24.72 13.13 13.05
Minimum 48.95 40.96 31.73 26.48
Maximum 117.70 118.20 73.29 66.39
Average profit margin (percentage)
Average 0.0844 0.1631 0.3 0.3 228.57
ÃÃ
28.28
ÃÃ
Median 0.0763 0.1369
Standard deviation 0.0489 0.1071
Minimum 20.0116 0.0260
Maximum 0.1699 0.4426
Premium (percentage)
Average 0.0315 0.0548 0.037248 0.014512 20.00635 23.7275
ÃÃ
Median 0.0192 0.0184 0.026786 0.008451
Standard deviation 0.0335 0.1358 0.033978 0.016395
Minimum 0.0001 0.0008 0.000558 0.000103
Maximum 0.1378 0.8592 0.135012 0.093025
Realised profit (percentage)
Average 0.067121 0.02004 0.735979 0.863739 0.422055 24.6E-06
Median 0.074466 0.07729 0.756618 0.891625
Standard deviation 0.106639 0.318446 0.111909 0.119258
Minimum 20.14145 20.46684 0.536689 0.629435
Maximum 0.293944 0.577464 0.895124 1.032362
ÃÃ
Statistically significant at the 5 per cent level.
AUCTIONS AND LAND VALUES 2251
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
Public Information (H2)
Does public information reduce overbidding?
The evidence in Table 2 suggests so. The
average profit margins are significantly higher
when public information is presented to the par-
ticipants. When public information in the form
of the lowest private signal is made known to
participants,
21
the mean premiums tend to be
Figure 1. Average profit margin.
Figure 2. Land prices (winning bids).
2252 CARL R. GWIN ET AL.
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
higher. While this evidence may appear
counter-intuitive, we note that premiums are
relative only to the second-highest bid. It says
nothing about the land price per se. So the
null in the second hypothesis is rejected.
As an aside, we compare the land price
(winning bid) between scenarios with and
without public information. The land prices
in scenario B are generally lower than those
in scenario A but not significantly so.
Figure 3. Actual (A) versus theoretical (T) realised profits.
Figure 4. Premiums.
AUCTIONS AND LAND VALUES 2253
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
Role of Experience (H3)
To address the issue of whether participants
learn over repeated auctions, we examine the
range of bids over time. Range is simply the
difference between the highest and lowest
bids for each period. We plot the range for
scenarios A and B in Figure 6.
Table 3 shows the regression analysis for
the different scenarios where the range in
each period is regressed on time t.
Range
t
¼ a þbt (6)
The null hypothesis is that if the coefficient
on t is not significant, then experience is
not an important factor for participants. The
coefficient on time (second column in
Table 3) is negative for all four scenarios. In
other words, the regression results show that
convergence in bids occurs where the range
narrows over time. However, the coefficient
on time is significant only for scenario
B. This evidence rejects the null hypothesis
that experience is not an important factor.
On the contrary, experience results in a
lower range over time. Collectively, the
results here underscore the importance of
public information in the learning process
for repeated bidders.
Table 2. Difference in means for effect of public information
Scenarios T-statistic
Average profit margin (percentage)
Scenario A – Scenario B 25.5115
ÃÃ
Premium (percentage)
Scenario A – Scenario B 22.1065
ÃÃ
Realised profit (percentage)
Scenario A – Scenario B 23.6 E-5
ÃÃ
statistically significant at the 5 per cent level.
Figure 5. Minimum profit margin.
2254 CARL R. GWIN ET AL.
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
How valuable is the private signal? To
answer this question, we examine the hit
ratio. As explained earlier, if participants
rely completely on their private signal and
conform to an accepted profit margin, the
winner in each round of bidding should be the
participant with the highest private signal.
Under these circumstances, the hit ratio will
be 100 per cent. Table A1 in the Appendix dis-
plays the incidences of hits. We note that the hit
ratio for scenario Ais 40.48 per cent, suggesting
that participants rely on their private signals.
Robustness Test
A question that arises from the research design
in this experiment concerns the heterogeneity
of the participants in the two scenarios
(A and B). In other words, to what extent, if
any, do the differences in the psychological
make-up of participants in the two groups
drive the results? To address this question, we
repeated scenario B with the group of partici-
pants from scenario A. Our tests show that
earlier results are not qualitatively affected.
What is different is that, since the repeat game
is held after all scenarios are over, the partici-
pants exhibited lower tendencies to overbid-
ding. We interpret this as evidence of the role
of experience.
22
6. Policy Implications and Conclusion
This paper is motivated by the lack of theoreti-
cal work and empirical evidence on the issue
of whether bidding experience in a repeated
game framework mitigates overbidding. We
focus our study on land auctions that expose
developers to price uncertainty by appealing
to an experimental methodology. The results
show that overbidding is pervasive even in
repeated rounds of bidding. Developers (par-
ticipants) tend to overbid by shaving their
expected profit margins relative to the theor-
etical benchmark bids based on a 30 per cent
Figure 6. Range of bids.
Table 3. Regression of range on time
Scenario Coefficient Standard error T-statistic
A 20.09196 0.073335 21.25392
B 20.18609
ÃÃ
0.090414 22.05817
ÃÃ
statistically significant at the 5 per cent level.
AUCTIONS AND LAND VALUES 2255
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
profit margin and the Risk Neutral Nash
Equilibrium (RNNE). Premiums of the
winning bid are higher than the theoretical
premiums. Bidders even accept negative
profit margins, exposing themselves to
greater price risk. The realised profits reflect
the tendency to overbid.
The policy implication of this result for
land development is clear. Where land is
made available for development via competi-
tive bidding or auctions, a likely consequence
would be that the value of land would be
inflated due to overbidding. Needless to say,
such aggressive bidding behaviour results in
significantly higher revenues for the land
supplier. So it would not be surprising to
find that auctions are the preferred mechanism
for land allocation in some countries.
Public information, defined narrowly in this
study as information about the lowest future
property value (private signal), significantly
reduces overbidding. While the role of the
lowest private signal is well documented in
the economics literature, the effects in land
auctions have not received sufficient attention.
The immediate implication is that a govern-
ment policy of making public information
about future property values is useful to
dampen aggressive bidding.
The experiments also indicate that bidders
do learn over time and that experience
matters. This is evidenced by the observation
that the range of bids tend to narrow over
time—i.e. convergence in bids occurs with
experience. However, this result is statisti-
cally significant only when public infor-
mation is available. The range of bids
shows a significant decrease in time when
public information is available. Here, the
policy implication is that countries that are
new to auctions as an allocation mechanism
for land development should allow time for
developers to gain experience in bidding.
Land prices determined at initial auctions
may be somewhat inflated. This factor
should be borne in mind when analysing
the values of land rights in China, for
instance. Policy-makers should also consider
the consequent impact on property prices
should developers be able to pass on the
higher land costs to end-consumers.
Several policy-related questions should be
highlighted. While land auctions generate sur-
pluses for governments, should the surpluses
be utilised to pay for future social costs associ-
ated with private ownership, or to recoup
infrastructure costs incurred with urban devel-
opment? Or should auction revenues be used
to cross-subsidise other government pro-
grammes? On a broader economic perspec-
tive, it would also be interesting to study the
general equilibrium outcomes of land auc-
tions. Would consistent overbidding in land
auctions divert capital from other productive
uses? How efficient is a land auction system
if developers can pass on higher land costs
to buyers/users in the form of higher prices
or rents?
As with all such experiments, we acknowl-
edge that the results are limited by the game
parameter values we chose. Future work
could test the robustness of the results under
different parameter values. The endogeneity
of the supply and price relation is largely sim-
plified in our analysis. Treating supply and
price as a system of equations is also another
interesting area of research. Finally, a major
assumption made in this study is the atomistic
nature of property developers. This assump-
tion may be violated in practice (Ong et al.,
2003). A related issue is to examine the
number of developers that constitute an oligo-
poly. The authors are currently pursuing
research into bidding in an oligopolistic
framework.
In summary, this paper focuses on the effect
on land values of land allocation using the
auction mechanism. The closed tender (first-
price sealed bid) mechanism is used in this
research, but future research could focus on
different auction mechanisms. Nonetheless,
we hope that the experimental approach
would gain greater acceptance in the analysis
of urban development issues.
Notes
1. In Hong Kong, land auctions are held regu-
larly under the open call system, while the
closed tender or first-price sealed bid
2256 CARL R. GWIN ET AL.
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
(FPSB) auction is used in Singapore.
However, the objective of this study is not
to compare various auction mechanisms for
land development.
2. Industry estimates place developers’ profit
margin as approximately 30 per cent to
cover the cost of capital and a risk
premium. Hadhi (1997) reports that develo-
pers’ margins have fallen below 10 per
cent.
3. Evidence of this would suggest collusion
among developers.
4. One extra person is designated as the alter-
nate, who enters the game only if a partici-
pant is declared bankrupt.
5. A pilot experiment was conducted with part-
time Masters students who were real estate
professionals with working experience. The
results are qualitatively unchanged.
6. To simulate business risk, participants can
be declared bankrupt if they do not have
sufficient capital and cash to meet obli-
gations such as bid price and construction
costs.
7. This assumption is made to eliminate private
values for land arising from developer-
specific differences in development costs. It
is worth noting that differences in cost can
lead to oligopolistic competition among
developers, an area of research the authors
are currently pursuing.
8. No pre-sale or marketing of units is allowed
before completion to avoid moral hazard and
adverse selection problems (see Ong, 1996
and 1999).
9. The private signals are based on a uniform
distribution centred on the property price
four periods ahead.
10. To maintain realism, a pre-specified time is
given for bids to be submitted. Participants
whofail tosubmit bids beforethe pre-specified
time limit will be excluded from that round.
11. Our study only incorporates profit maximi-
sation in the objective function. Other con-
siderations such as financial leverage, tax,
portfolio and reputation considerations are
ignored.
12. Friedman and Sunder (1994) find that class
grades rank a very distant second to cash
rewards.
13. There are three conditions sufficient to
induce participants’ characteristics: mono-
tonicity—subjects must prefer more
reward to less; salience—the reward
received must depend on the subject’s
actions; and, dominance—subject’s utility
from the experiment comes predominantly
from the reward medium and other influ-
ences are negligible (Friedman and
Sunder, 1994). Cash compensation satisfies
all three conditions.
14. The lottery game goes like this. There is an
electronic number generator that will ran-
domly generate a number ranging from 1
to 100. If the number generated is
between 1 and 70 (inclusive), the partici-
pant wins $100. However, if the number
generated is between 71 and 100 (inclus-
ive), the participant wins nothing. In other
words, the participant has a 70 per cent
chance of winning $100 in this lottery.
But before playing the lottery game, a
buyer makes the participant an offer to
buy the right to play the game. The buyer
has in mind an undisclosed offer price z,
which is no more than $70. However, the
buyer wants the participant to name the
price at which he is willing to sell the
right to play the game. After naming the
selling price, the buyer will reveal the
offer price z. If the selling price is less
than z, the participant will get z. If the
selling price is more than or equal to z,
then the participant’s price is higher than
what the buyer is willing to pay and the
participant will proceed to play the lottery.
15. The 30 per cent benchmark margin is chosen
arbitrarily, based on industry observations.
Although participants were informed of
this industry benchmark, they could choose
any profit margin. One concern was that
the experiment does not provide alternative
investment commensurate with the 30 per
cent margin for real estate development.
However, participants are rewarded at the
end of the game based on their relative
cumulative profits (or losses). Thus, the
reward system in this experiment should
enable bidders to trade-off expected returns
from bidding and the associated risk.
Bidders that were too aggressive and had
very low ex ante profit expectations could
end up incurring losses, or lower profits
relative to other participants.
16. We use the lowest value as opposed to an
expected value to conform with prior
research in the economics literature.
17. We observed that some participants are
conservative in their bidding behaviour—i.e.
they exhibit strategic bid shaving behaviour.
18. Alternative benchmark profit margins
ranging from 10–25 per cent provide
qualitatively similar results.
19. The minimum expected profit margin is also
indicative of errors in human judgements or
calculations.
20. This is because the winning bid may be the
participant with the highest private signal.
AUCTIONS AND LAND VALUES 2257
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
21. Participants in a survey after the experiment
reveal that they place high weights on the
public signal.
22. To avoid a cohort effect, the experiment was
repeated on an entirely different and new
group of participants. Similar results were
obtained.
References
ADAMS, P., KLUGER, B. and WYATT, S. (1992)
Integrating auction and search markets: the
slow Dutch auction, Journal of Real Estate
Finance and Economics, 5(3), pp. 239–253.
ALLEN, M. T. and SWISHER, J. (2000) An analysis
of the price formation process at a HUD auction,
Journal of Real Estate Research, 20(3), pp.
279–298.
COLWELL, P. and YAVAS, A. (1994) The demand
for agricultural land and strategic bidding in
auctions, Journal of Real Estate Finance and
Economics, 8(2), pp. 137–149.
DOTZOUR, M. G., MOORHEAD, E. andWINKLER, D. T.
(1998) The impact of auctions on residential
sales prices in New Zealand, Journal of Real
Estate Research, 16(1), pp. 57–71.
DYER, D., KAGEL, J. and LEVIN, D. (1989) A com-
parison of naı ¨ve and experienced bidders in
common value offer auctions: a laboratory
analysis, Economic Journal, 99, pp. 108–115.
FIORINA, M. and PLOTT, C. (1978) Committee
decisions under majority rule: an experimental
study, American Political Science Review, 72,
pp. 575–598.
FRIEDMAN, D. (1993) Howtrading institutions affect
financial market performance: some laboratory
evidence, Economic Inquiry, 31(3), pp. 410– 435.
FRIEDMAN, D. and SUNDER, S. (1994) Experimen-
tal Methods: A Primer for Economists.
New York: Cambridge University Press.
FU, Y. and CHING, S. (2003) The contestability of
the urban land market: an event study of Hong
Kong Land auctions, Regional Science and
Urban Economics, 33(6), pp. 695–720.
GOEREE, J. K. and OFFERMAN, T. (2002) Efficiency
in auctions with private and common values: an
experimental study, American Economic
Review, 92(3), pp. 625–643.
GRESIK, T. andSATTERTHWAITE, M. (1986) The rate
at which a simple market becomes efficient as the
number of traders increases: an asymptotic result
for optimal trading mechanism. October.
CMSEMS DP No. 708, Northwestern University.
HADHI, A. (1997) Developers’ margins fall below
10%, BTOnline, 18 November.
HANSEN, R. and LOTT, R. J. Jr. (1991) The
winner’s curse and public information in
common value auctions: comment, American
Economic Review, 81, pp. 347–361.
JAMAL, K. and SUNDER, S. (1991) Money vs.
gaming: effects of salient monetary payments in
double oral auctions, Organizational Behavior
and Human Decision Processes, 49, pp. 151–166.
KAGEL, J. and LEVIN, D. (1986) The winner’s curse
and public information in common value auctions,
American Economic Review, 76, pp. 894–920.
KAGEL, J. and LEVIN, D. (1991) The winner’s
curse and public information in common value
auctions: reply, American Economic Review,
81, pp. 362–369.
KAGEL, J. H. and RICHARD, J. F. (2001)
Super-experienced bidders in first-price
common-value auctions: rules of thumb, Nash
equilibrium bidding, and the winner’s curse,
Review of Economics and Statistics, 83(3),
pp. 408–419.
LIND, B. and PLOTT, C. (1991) The winner’s curse:
experiments with buyers and with sellers,
American Economic Review, 81, pp. 335–346.
LUSHT, K. (1996) A comparison of prices bought
by English auctions and private negotiations,
Real Estate Economics, 24(4), pp. 517–530.
MAYER, C. (1995) A model of negotiated sales
applied to real estate auctions, Journal of
Urban Economics, 38(1), pp. 1–22.
MAYER, C. (1998) Assessing the performance of
real estate auctions, Real Estate Economics,
26(1), pp. 41–66.
MCCABE, K., RASSENTI, S. and SMITH, V. (1993)
Designing a uniform price double auction: an
experimental evaluation, in: D. FRIEDMAN and
J. RUST (Eds) The Double Auction Market: Insti-
tutions, Theories and Evidence, pp. 307–332.
Santa Fe Institute Series in the Science of the
Complexity, Proceedings, Vol. 15, Reading,
MA: Addis.
MILGROM, P. (2004) Putting Auction Theory
to Work. Cambridge: Cambridge University
Press.
MILGROM, P. and WEBER, R. (1982) A theory of
auctions and competitive bidding, Econome-
trica, 50, pp. 1485–1527.
ONG, S. E. (1996) Building defects, warranties and
project financing from pre-completion market-
ing, Journal of Property Finance, 8(1), pp.
35–51.
ONG, S. E. (1999) Caveat emptor: adverse selec-
tion in buying properties under construction,
Property Management, 17(1) pp. 49–64.
ONG, S. E., CHENG, F. J., BOON, B. and SING, T. F.
(2003) Oligopolistic bidding and pricing in real
estate development: experimental evidence,
Journal of Property Investment and Finance,
21(2) pp. 154–189.
OOI, J., SIRMANS, C. F. and TURNBULL, G. K.
(2005) Price formation under small numbers
competition: evidence from land auctions
in Singapore, Real Estate Economics
(forthcoming).
2258 CARL R. GWIN ET AL.
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from
PALFREY, T. and PORTER, R. (1991) Guidelines for
submission of manuscripts on experimental econ-
omics, Econometrica, 59(4), pp. 1197–1198.
PLOTT, C. (1982) Industrial organisation theory
and experimental economics, Journal of Econ-
omic Literature, 20, pp. 1485–1527.
PLOTT, C. (1987) Dimensions of parallelism:
some policy applications of experimental
methods, in: A. ROTH (Ed.) Laboratory Exper-
imentation in Economics, Six Points of View,
pp. 193–219. Cambridge: Cambridge Univer-
sity Press.
QUAN, D. (1994) Real estate auctions: a survey of
theory and practice, Journal of Real Estate
Finance and Economics, 9(1), pp. 23–49.
ROTH, A. and MALOUF, M. (1979) Game theoretic
models and the role of information in bargain-
ing: an experimental study, Psychological
Review, 86, pp. 574–594.
SAMUELSON, W. and BAZERMAN, M. (1985) The
winner’s curse in bilateral negotiations, in:
V. SMITH (Ed.) Research in Experimental Econ-
omics, Vol 3, pp. 105–137. Greenwich, CT: JAI
Press.
SATTERTHWAITE, M. and WILLIAMS, S. (1993)
The Bayesian theory of the k-double auction,
in: D. FRIEDMAN and J. RUST (Eds) The
Double Auction Market: Institutions, Theories
and Evidence, pp. 99–124. Santa Fe Institute
Series in the Science of the Complexity, Pro-
ceedings, Vol. 15, Reading, MA: Addis.
SMITH, V. and WALKER, J. (1992) Rewards,
experience and decision costs in first price auc-
tions, Economic Inquiry, 32(1), pp. 138–145.
TANG, Y. (1989) Urban land use in China, Land
Use Policy, 6(1), pp. 53–63.
VANDERPORTEN, B. (1992) Timing of bids at pooled
real estate auctions, Journal of Real Estate
Finance and Economics, 5(3), pp. 255–267.
WILSON, R. (1977) A bidding model of perfect
competition, Review of Economic Studies, 44,
pp. 511–518.
WU, F. and YEH, A. (1997) Changing spatial distri-
bution and determinants of land development in
Chinese cities in the transition from a centrally
planned economy to a socialist market
economy: a case study of Guangzhou, Urban
Studies, 34(11), pp. 1851–1879.
XIE, Q., PARSA, A. and REDDING, B. (2002) The
emergence of the urban land market in China:
evolution, structure, constraints and perspec-
tives, Urban Studies, 39(8), pp. 1375–1398.
YAVAS, A., ABRAMS, E. and SEFTON, M. (2000)
An experimental comparison of two search
models, Economic Theory, 16, pp. 735–749.
YAVAS, A., MICELI, T. and SIRMANS, C. F. (2001)
An experimental analysis of the impact of inter-
mediaries on the outcome of bargaining games,
Real Estate Economics, 29, pp. 61–86.
Appendix
Table A1. Number of hits and hit ratio
t Scenario AScenario B
1
2
3
4 1
5 1
6 1
7 1
8 1
9
10
11 1
12
13 1
14 1
15
16 1
17 1
18
19
20 1 1
21 1
22 1
23 1
24 1
25
26 1
27 1
28 1
29
30 1
31 1
32
33 1
34
35
36 1 1
37
38
39 1 1
40 1
41
42
Total hits 17 10
Hit ratio 40.48 25.00
AUCTIONS AND LAND VALUES 2259
at INDIAN INST MGMNT BANGALORE on June 19, 2013 usj.sagepub.com Downloaded from

Sponsor Documents

Or use your account on DocShare.tips

Hide

Forgot your password?

Or register your new account on DocShare.tips

Hide

Lost your password? Please enter your email address. You will receive a link to create a new password.

Back to log-in

Close