Price Discovery in Real Estate Auctions:
The Story of Unsuccessful Attempts
Seow Eng ONG
Department of Real Estate
National University of Singapore
March 26, 2004
Abstract
Little is known of the effects of the auction mechanism in relation to postauction market sale. This empirical study of 966 unsuccessful auction observations
over a period of almost five years is an initial attempt to provide insights on post
auction effects. Approximately half of these properties were eventually sold via
private negotiations. The post-auction privately negotiated sale prices are on average
higher than the opening and last bids when at least one bid is received at the auction.
However, properties that received no bids at the auction were transacted on average, at
a discount to the opening bids. The probability of a subsequent post-auction
transaction is significantly higher for apartments and terrace houses and when auction
turnout is high; and lower in the absence of any bid and in some years. The price to
opening bid differential increases with the number of increments at the auction, but is
lower for more atypical properties and decreases with time-to-subsequent sale. The
results lend support to the hypothesis that the auction mechanism provides a positive
price discovery effect, ostensibly from publicity and exposure at the auctions.
Interestingly, the probability of subsequent sale is also decreasing (albeit
insignificantly) in the capital loss experienced by owners, lending some support to the
loss aversion theory. However, the attenuation in opening bids is observed for all
properties, regardless of whether they are put up for sale by owners or by financial
institutions.
Acknowledgements:
The author wishes to thank Peter Colwell, Ken Lusht and Paul Anglin for their
insightful comments, and Chee-Yong Mak for excellent research assistance. Funding
from NUS is gratefully acknowledged. Please email comments to seong@nus.edu.sg.
1
Price Discovery in Real Estate Auctions:
The Story of Unsuccessful Attempts
1.
Introduction
An auction is a market institution with an explicit set of rules determining
resource allocation and prices on the basis of bids from market participants (McAfee
and MacMillan, 1987). Auction theory provides one explicit but restricted model of
price formation in that it ignores the bargaining aspects of the process (Arrow, 1959).
Extant researches in US, Australia, New Zealand and Germany have emphasized on
the comparison of this selling methods via different auction variations1 and/or private
negotiations, and the factors determining auction prices. The expected prices for
auctioned properties are often derived using hedonic price models and compared to
realized prices, not withstanding the selection bias limitation of the hedonic function
(Mayer, 1998).
The process of price discovery during and after the auctions, in particular the
effects on post-auction prices and multiple auctions attempts, have been somewhat
neglected in the literature. Ashenfelter and Genesove (1992) is one of the few papers
that examined post-auction price behavior. They show that the auction prices for
identical condominium units in New Jersey were 13% higher than the prices
determined in subsequent face-to-face bargaining. This arose because of the unique
institutional framework where the “hammer price” is not necessarily binding,2 and the
original auction sale could “fall through”. This paper complements the earlier work by
using data from an institutional setup that binds “hammer prices” and auction sales.
Under this alternative framework, it is clear that only genuine bidders participate at the
auctions. However, unsuccessful sellers could seek out interested buyers after the
auction and privately negotiate.
This paper focuses on post-auction privately negotiated sales. By so doing, we
attempt to better understand the price discovery process after unsuccessful auction
1
English ascending bid auctions, Dutch auctions, First-price sealed-bid auctions and Second-price
sealed-bid auctions.
2
37% of the sample in the Ashenfelter and Genesove (1992) study were not sold at the “hammer
prices”.
2
attempts and bridge the gap between auctions and negotiated sales. Hitherto, these two
processes are viewed as mutually exclusive modes of price discovery (Quan, 2002;
Allen and Swisher, 2000; Mayer 1998; Dotzour, Moorhead and Winkler, 1998).
The process of being put up for auction could provide valuable exposure for
the property. Even though the auction was unsuccessful, the auction process could be
used to identify potential buyers, and as such, could be viewed as an alternative or
augmentation to the listing process. The main objective of this paper will be to
examine the determinants that influence post-auction sales, the deviation in eventual
sale price from opening bids and the time before private sale is concluded. In addition,
we study the interactive influence of market interest such as turnout, number of bids
(or no-bids), auction houses, property type, state of the market and seller motivation as
proxied by the number of previous attempts and distress versus non-distress sales.
Publicity exposure is deemed to have a positive price discovery effect if (a) a
subsequent privately negotiated sale occurs and (b) the privately negotiated sale price
is higher than some benchmark values determined at the time of the auction. For this
study, we use the opening bid and last bid before the property is withdrawn from
auction as the benchmark values to compute the price-bid differentials. In addition, we
examine how the price-bid differentials change with the time taken to successfully
negotiate a sale.
Reserve prices would have been preferred benchmark values, but are
unavailable in our dataset. Although we estimated the market values for auctioned
properties, they are at best noisy estimates given uncertain market conditions,
especially in a market downturn where transaction volumes are low. We opted for
opening bids as these provide an indication of seller information set and motivation.
This paper is closely related in spirit to the work by Knight (2002) and Anglin
(2004) where the focus is on changes in listing prices and re-listings. This work differs
not only by focusing on auctions rather than MLS sales, we focus on the opening bids,
which are non-binding lower bounds rather than listing prices, which are upper bounds
set by sellers.
A secondary objective for this paper is to provide an alternative test of the loss
aversion theory advocated by Genesove and Mayer (2001). The loss aversion theory
3
postulates that owners who are loss-averse have an incentive to attenuate that loss by
choosing a reservation price that exceeds the level they would set in the absence of a
loss. By extension, it can be hypothesized that owners seek to avoid loss realization if
the expected loss is too high. If so, then the probability of a subsequent sale should be
decreasing in the expected loss. However, this should only be true for owner-initiated
sale rather than distress sale auction.
The dataset is a comprehensive coverage of all residential auctions in
Singapore from 1995 to 2000. Approximately half of 966 unsuccessfully auctioned
properties were eventually sold via private negotiations. The probability of a
subsequent post-auction transaction is significantly higher for apartments and terrace
houses and when auction turnout is high; and lower in the absence of any bid and in
some years. The price to opening bid differential increases with the number of
increments at the auction, but is lower for more atypical properties and decreases with
time-to-subsequent sale.
The results lend support to the hypothesis that the auction mechanism provides
a positive price discovery effect, ostensibly from publicity and exposure at the
auctions. This suggests that sellers could benefit from putting up their property for
auction at least once even though it may not be successfully sold at the action. At the
minimum, the auction process would provide a gauge of market interest and could
increase the likelihood and price of subsequent privately negotiated transactions.
We find that the probability of subsequent sale is also decreasing (albeit
insignificantly) in the capital loss experienced by owners, lending some support to the
loss aversion theory advocated by Genesove and Mayer (2001). However, the
attenuation in opening bids is observed for all properties, regardless of whether they
are put up for sale by owners or by financial institutions.
This paper will be organized into a five sections with a brief outline of the past
researches featured in the section 2. A brief discussion of the auction market in
Singapore will be presented in section 3. Section 4 discusses the data used for this
empirical study while section 5 presents the findings and analysis. Section 6
concludes.
4
2.
Literature Review
Ashenfelter and Genesove (1992) is one of the earlier studies on real estate
auctions where they show that the prices for identical units were 13% higher than for
units subsequently sold in private negotiations. Other researches have centralized on
the changes in the auction prices as the auction proceeds. Lusht (1994) and
Vanderporten (1992) studied and quantified the extent of the ‘afternoon effect’ in
which prices gradually moves downwards during a single auction process of multiple
objects. Ashenfelter and Genesove (1992) reported a 0.27% decline per unit sold or
about 10% from the first object sold to the last, during an auction of 83 similar
condominiums in New Jersey. Lusht (1994) presented similar findings with regard to a
one-time auction process of an Australian commercial bank’s 50 branches. Quality
differences are controlled as each of the branches had similar investment attributes and
were sold with the same terms. The discount between the first and the last object
auctioned was approximately 8%. Comparisons of differences in expected and private
market transaction prices of post-auction properties were not performed.
Mayer (1998) showed that a quick sale results in a poorer “match” between the
buyer and the house, thus resulting in a discount. A poorer “match” will translate into
a higher “mismatch” cost compared to a private negotiated sale that would allow more
time for the buyer to search for his ideal home. Conversely, Lusht (1996) found that
auctions would yield higher prices than privately negotiated sales. Newell,
MacFarlane, Lusht and Bulloch (1993) studied the housing market in Sydney,
Australia and found that the median sales price of auctioned properties was 3.6%
higher than the median sales price of properties sold by private treaty. Quality
differences in the study are not accounted for.
Lusht (1996) compared the sales prices of houses in Melbourne and found that
auctions of middle to high priced homes produce about 8% price premium. Bids in
auctions may provide an attractive alternative to estimating the value of a specific
property, particularly in periods when distress sales are rampant and auctions are
frequent (Quan, 1994). He postulates that the auction theory relates the beliefs and the
private information of individual bidders to their bids even if there are few bidders.
Inference of the valuation of the highest bidder in a seal-bid auction corresponds much
5
more closely to the highest and best use definition of market value, even if there are
few bidders.
Having established the informational value of auction bids, comparison of
these bids with post-auction private transaction prices of the same properties will
enable us to discover the likely effects of the auction process on the latter. This will be
discussed in a later section.
3.
Auctions in Singapore
The dominant auction format in Singapore is the English ascending bid
auction. Auction has generally been regarded as a last resort method of disposal. The
local sentiment towards auction is similar to that of the US, where auctions are
associated with distress properties – foreclosure or mortgagee sales (Asabere and
Huffman, 1992). Distress sales are typically put up by the mortgagee (banks/financial
institution). There was a surge in auction sales following the Asian financial crisis
from 1998-1999. Although a good proportion comprises mortgagee sales, there has
been a discernible increase in owner auctions, due to a diminution of the stigma
associated with auctions. Even so, the number of properties put up for auction is very
low, and reached a high of approximately 3.5% of the total number of property
transactions in 1998 and 1999. There has also been a growing perception among
potential buyers that auctions of distress properties provide a good avenue to acquire
properties at bargain prices. Buyers and sellers have better understanding and
awareness of the efficiency of auction system as a method of sale, and auction
companies in Singapore have substantially increased the frequency of auctions held
each month to meet the growing demand.
Bidders in Singapore are generally not aggressive. The success rate for each
bidding session varies from 10% to 50%; while the percentage of successful postauction private market transaction is higher. The low success rate was attributed to the
flagging performance in general, rather than the appeal of auction itself.3 Another
possible cause of the low success rates is that many owners use the auction process as
a gauge of market interest in their properties. In addition, some buyers withhold from
3
Fang, N. (2000, November 1). Slow off the block. The Straits Times.
6
biding during an auction in the hope of securing lower transaction prices in postauction private negotiations. Overly conservative bidding and the expectation that
private negotiations are more likely to secure sale may induce sellers to set
unrealistically high reserve prices.
4.
Data
Our sample comprises 1,654 private residential properties that are being put up
for auction from 3Q1995 to 1Q2000. This sample covers all residential auctions over
that period. Residential properties are typically classified into high-rise (apartment and
condominium) and low-rise (terrace, semi-detached, detached houses). The data set
includes variables on the location, date of auction, number of turnout, auctioneer,
distress sale versus by-owner sale, type of property, tenure, opening bid, last bid, and
number of bid increments during the auction. Properties that were put up for more than
one attempt were omitted, reducing the sample size to 1,292 properties. Of these, 326
properties were sold at the auction. The remaining 966 properties that were not sold at
the auction form the focus of our analysis.
Property transaction data are extracted from the ReaLink for a period from
1975 Q1 to 2000 Q4, compiled by the Singapore Institute of Surveyor and Valuers
(SISV). We compute an estimated market value for all properties put up for auction
using either the last transacted price or comparable sales of other private residential
properties that are of the same type and in the same estate, having very similar floor
area and floor level.
We develop a proxy measure called the level of reserve price (LRP) (DeBoer,
et. al, 1992), which is defined as the difference between the property’s estimated
market value and the opening bid, divided by the opening bid. The estimated market
value is the last transaction price for the auctioned property or the estimated market
value of properties most comparable to the auctioned property if there is no previous
transaction, adjusted by the price index (RPI). Our adjustment uses the RPI from the
quarter prior to the auction, as information on the contemporaneous price index is not
available at the time of auction.
7
The state of the market (SOM) indirectly affects the sentiment of property
buyers and hence affects the probability of a sale (Mayer, 1995). SOM is a dummy
variable given a value of zero if the auction occurred in a quarter following two
previous successive quarters of negative growth in the RPI. We also introduce year
dummy variables to control for the timing of the auction (Vanderporten, 1992), and
property type dummy variable to distinguish high-rise properties from low-rise
properties – terrace, semi-detached and detached houses (TERR, SEMID and
DETACH). The data set also allows us to identify properties that were auctioned with
vacant possession (VP) and that were distress sales (DISTR).
Control dummy variables for the auction houses are also created for the big
four companies – Knight Frank (KF), Colliers Jardine (CJ), Jones Lang LaSalle (JLL),
and DTZ Debenham Tie Leung (DTZ). Maher (1989) and Ong, et. al (2003) show that
the auction house may contribute to the probability of sale. In
addition,
we
have
information on the turnout (TURNOUT) at the auctions (Burns, 1985) as well as the
number of increments (INCREM) during the bidding process (Ching and Fu, 2003).
In addition, we tracked the transactions for all the unsuccessful properties
(until end 2000). There were a total of 456 post-auction transactions, indicating that
close to half of the properties that were not sold at auctions were subsequently sold
through private negotiations. Of these, 122 properties received at least one bid during
the auction, and the remaining 334 properties did not receive any bid. The post-auction
transactions prices of the 456 properties are used to compute the price-bid differentials
using the opening bid and last bid before the property was withdrawn (TPOB and
TPSB, respectively). The time between the subsequent sale and auction dates is the
time-to-(subsequent)-sale (TOS).
To evaluate the loss aversion theory, we compute the expected loss using the
difference between the estimated market value and the original purchase price.
However, the original purchase price is available for only 583 properties in our
sample. As in Genesove and Mayer (2001), we define LOSS as the greater of the
difference between estimated market value and original purchase price, and zero.
Since loss aversion is attributable only to owner-sellers, we create an interaction term
(LOSSOWN) between LOSS and an owner dummy variable.
8
Exhibit 1 provides the description and summary statistics for variables defined
for this study.
5.
Analysis
Price-Bid Differentials
The preliminary evidence shows that properties that garnered some interest
(bids) sell (when they do) at higher prices than the opening bids and in a shorter time
compared to properties that received no bid. We observe that the average price-open
bid differential (TPOB) is -2.01% for the sample of 456 properties that were
subsequently sold. However, if we segregate these into properties that received at least
one bid (category 1) from those that received no bid (category 2), the average TPOBs
are 3.22% and -3.92% respectively. The difference is statistically different at 5% level
(t-stat = 4.379). See Exhibit 2a. The average time-to-sale (TOS) of 111 days and 131
days for the two categories of properties, respectively (the difference is also
statistically different at 5% level; t-stat = 2.7401).
In addition, the average price-last bid differential (TPSB) for the 122
properties that were eventually sold is 0.65%. Although this is not statistically
different from zero, it shows that some properties were negotiated at higher prices than
the last bid at the auction.
Exhibits 2b and 3 show the average price-open bid differentials (TPOB) for the
two categories by TOS. Positive average price-open bid differentials are observed for
properties that attracted some interest and that were subsequently sold within 180 days
of the auction. In contrast, properties that received no bid were sold at discounts to the
opening bids regardless of the time-to-sale. Positive
average
price-last
bid
differentials (TPSB) were also observed for properties sold within 180 days (Exhibits
2c and 4).
Although the average price-open bid differential (TPOB) for the category one
properties is statistically lower than the average price-open bid differential for
properties that were sold at the auctions (7.78%), it is interesting to observe that the
average TPOB for category one properties sold within 60 to 90 days is in fact higher
9
(9.27%). The average TPOB for properties sold within 180 days is 5.14%.4 This
evidence suggests that a positive price discovery effect occurred for properties that
attracted market interest at the auction.
Probability of Subsequent Privately Negotiated Sale
A probit model is estimated for the likelihood of a subsequent privately
negotiated sale (SUBOUT=1). We adopt essentially the same set of explanatory
variables that are used to examine the factors that influence sale at the auction. The
results are given in Exhibit 5 (model 1). Properties that received no bid (NOBID=1)
have a lower probability of sale (p-value=0.0675), supporting the earlier results. The
turnout at the auction (TURNOUT), however, is significant and positive. This result is
also consistent with our earlier study (Ong, et. al, 2003). Surprisingly, distress sale, the
level of reserve price, state of market, number of increments and number of previous
attempts are not statistically significant.
Not surprisingly, the probability of a subsequent sale is reduced in 1999 and
2000, when the market is recovering. This suggests that post-auction price discovery is
likely to be more useful under uncertain market conditions. Interestingly, some auction
houses (KF and CJ) have a negative and significant effect on the probability of a
subsequent sale. These two auction houses, in fact, have a positive effect on the
probability of an auction sale (Ong, et. al, 2003). Taken together, the evidence
suggests that some tradeoff between at- and post-auction success rates for these
auction houses.
Exhibit 5 model 2 estimates the sub-sample where data on the original
purchase price is available. Unfortunately, original purchase prices are available only
for 583 observations, thus reducing the sample size. Two additional variables were
introduced – LOSS and LOSSOWN. As noted before, the loss aversion theory should
only hold for owner-initiated sale rather than distress sale auction since financial
institutions that are selling the defaulted properties are typically risk-neutral. By
extension, it can be hypothesized that owners seek to avoid loss realization if the loss
experienced is too high.
4
This is statistically different from the mean TPOB for auctioned properties (t-stat = 2.498).
10
It is observed that the coefficient on LOSS is positive (but insignificant),
indicating that the probability of a subsequent sale is increasing in expected loss. This
result is intuitive as the higher the loss, the more likely the seller would be keen to sell.
Interestingly, the coefficient on LOSSOWN is negative (p-value of 0.1288),
suggesting that owner-sellers are less likely to agree to a subsequent sale when the
expected loss is higher. This is weak evidence in support of the loss aversion theory.
Determinants of Price-Open Bid Differential
To examine how the price-bid differentials are influenced by various
explanatory variables, we estimate a simple regression model (Exhibit 6). The sample
selection issue (for properties that were subsequently sold) is controlled for by way of
estimating the Mills ratio from the probit model from Exhibit 5. Model 1 in Exhibit 6
reports the results only for properties that were subsequently transacted, while model 2
is estimated for the full sample to provide a comparison. The results show that the
price-open bid differential increases in the number of increments and turnout (model
2). We can interpret this result to imply that the eventual sale price (relative to opening
bid) is increasing in market interest, as proxied by turnout and bid increments.
The negative and significant coefficient on time-to-subsequent sale (TOS) is
also consistent with our earlier result showing that the price-bid differential is higher
when the sale is negotiated closer to the auction date. Lower differentials are observed
for more atypical properties such as terrace, semi-detached and detached houses,
although the reduction is statistically significant only for semi-detached houses.
The evidence on the number of previous attempts and no-bid is rather mixed.
The coefficient on the number of previous attempts is positive but insignificant in
model 1, suggesting that repeated attempts may help; but flipped to a negative sign in
model 2. No-bid is negative and significant, as expected, only in model 2.
Loss Aversion and Opening Bids
Genesove and Mayer (2001) tested for loss aversion in listing prices where
owners attenuate that loss by setting higher reservation prices. Our data allows us to
examine this theory in an alternative auction setting where the opening bid,
11
determined by the seller, provide a lower bound. The listing price, in contrast, is an
upper bound. Loss aversion would suggest that owner-sellers set higher opening bids
for higher expected losses.
We regress the opening bid on LOSS and LOSSOWN, controlling for other
pertinent variables such as year of auction, property type, distress sale, and market
condition (the price index at the time of auction is used as a proxy). Exhibit 7 shows
that the coefficient on LOSS is positive and significant, indicating that sellers set
higher opening bids to attenuate losses. However, this effect is insignificant for ownersellers. The coefficient on LOSSOWN, when regressed without the LOSS variable, is
positive but insignificant. The evidence here is consistent with the loss aversion
theory, although loss attenuation is observed for distress properties as well.
6.
Conclusion
This paper presents a preliminary examination of post-auction price discovery.
The premise is that the auction process itself may provide positive publicity or
exposure effects to the extent that sellers gather useful information even if no sale is
made at the auction. About half of the properties that were unsuccessful at the auction
were eventually sold via private negotiation. The majority of subsequent sales occur
within 180 days of the auction.
In contrast to Ashenfelter and Genesove (1992), we find that properties that
were subsequently sold through private negotiations sell on average at a premium to
opening prices.5 The price differential is positive for properties that attract some
market interest as evidenced by receiving at least one bid, as opposed to a negative
differential for properties that did not receive any bid.
The evidence is consistent with the market interest explanation – the higher the
level of market interest, the more likely a subsequent private sale and the higher the
price relative to opening bid. Interestingly, the number of increments received during
the auction affects only the price differential and not the probability of subsequent
sale. The results suggest that sellers could benefit from putting up their property for
auction at least once even though it may not be successfully sold at the action. At the
5
Unfortunately, we cannot do a direct comparison with Ashenfelter and Genesove (1992) as the
opening bids are not available in that dataset.
12
minimum, the auction process would provide a gauge of market interest and could
increase the likelihood and price of subsequent privately negotiated transactions.
By appealing to the observation that opening bids are lower bounds and that
loss-averse sellers are less likely to sell if the capital loss is too high, we are able to
conduct alternative tests of the loss aversion theory. We document weak evidence to
support the loss aversion theory in that a lower probability of subsequent sale is
observed for owner-sellers with higher losses. Attenuation in opening bids, however,
is observed for all properties, regardless of whether they are put up for sale by owners
or by financial institutions.
As previously mentioned, this paper is a preliminary attempt to better
understand the price discovery process that auctions provide. Many questions remain
unanswered. For example, the price differential for properties that were subsequently
sold is lower on average to that for auctioned properties, suggesting that (a) the
process of auction generate higher realized transaction prices as auction theory would
indicate (i.e., the winner’s curse) and/or (b) the opening bids for auctioned properties
are lower in comparison to properties that were not sold at the auction. Unraveling
these effects could be interesting to evaluate the information value of opening bids.
Also, how do prices for properties that were subsequently sold after the
auction compare against those that were not put up for auction? While we document
that properties sold at auctions has higher price-open bid differentials compared to
post-auction properties,6 are post-auction prices significantly different from auction
prices? Do the changes in opening bids in repeat attempts convey pertinent
information that could affect the probability of subsequent sale? Future work will seek
to address some of these questions.
6
This result could be attributed to differences in opening bids.
13
References
Allen, M. T. and J. Swisher, (2000) An Analysis of the Price Formation Process at a
HUD Auction, Journal of Real Estate Research, 20(3), 279-98.
Anglin, P., (2004) Selling and the Market Process: If, at First, You Don’t Succeed,
Try, Try, Try Again, AREUEA conference, Washington DC.
Asabere P. K. and Huffman, F. E. (1992) Price Determinants and Foreclosed Urban
Land, Urban Studies 29(5), 701-07.
Ashenfelter, O. and Genesove, D., (1992) Testing for Price Anomalies in Real Estate
Auction, American Economic Review, 501-05.
Burns, P., (1985) Market Structure and Buyer Behaviour: Price Adjustment in a Multiobject Progressive Oral Auction, Journal of Economic Behavior and Organization (6),
275-300.
Ching, S. and Fu, Y., (2003) Contestability of Urban Land Market: An Event Study of
Government Land Auctions in Hong Kong, Regional Science and Urban Economics,
33(6), 695 – 720.
DeBoer. L, Conrad. J, and Mcnamara, K., (1992) Property Tax Auction Sales, Land
Economics, 72-82.
Dotzour, M.G., Moorhead, E. and Winkler, D.T., (1998) Auctions on Residential Sales
Prices in New Zealand, Journal of Real Estate Research, 16:1, 57-70.
Genesove, D. and Mayer, C.J., (2001) Loss aversion and seller behavior: evidence
from the housing market, Quarterly Journal of Economics, 116, 1233 – 1260.
Knight, J.R., (2002) Listing price, Tom and ultimate selling price : cause and efforts of
listing price changes, Real Estate Economics, 30, 213—237
Lusht, K.M., (1994) Order and Price in a Sequential Auction. Journal of Real Estate
Finance and Economics, 8(3), 259 – 266.
Lusht, K.M., (1996) A Comparison of Prices Brought by English Auctions and Private
Negotiations, Real Estate Economics, 24, 517-530.
Maher, C., (1989) Information Intermediaries and Sales Strategy in an Urban Housing
Market: The Implications of Real Estate Auction in Melbourne, Urban Studies 26(5),
495-509.
Mayer, C.J., (1995) A Model of Negotiated Sales applied to Real-Estate Auctions,
Journal of Urban Economics 38 (1): 1-22.
14
Mayer, C. J., (1998) Assessing the Performance of Real Estate Auction, Real Estate
Economics, 26(1), 41-66.
McAfee, P. and McMillan, J., (1987) Auctions with a Stochastic Number of Bidders,
Journal of Economic Theory, 43, 1 – 19.
Newell, G., J. MacFarlane, K. Lusht and S. Bulloch., (1993) Empirical Analysis of
Real Estate Auction Versus Private Sales Performance. Working Paper, University of
Western Sydney: Hawkesbury, Australia.
Ong, S. E., Lusht, K., Mak, C. Y., (2003) Determinants of Successful Real Estate
Auction: Evidence from Singapore, AREUEA conference, Washington DC.
Quan, D.C., (1994) Real Estate Auctions: A Survey of Theory and Practice, Journal of
Real Estate Finance and Economics, 23-49.
Quan, D.C., (2002) Market mechanism choice and real estate disposition: Search
versus auction, Real Estate Economics 30 (3): 365-384.
Vanderporten, B., (1992) Timing of Bids at Pooled Real Estate Auctions, Journal of
Real Estate Finance and Economics, 255-67.
15
Exhibit 1a: Variables for Auction Data (3Q1995 – 1Q2000)
Variable
Description
DISTR
Distress Sale = 1
LRP
Level of reserve price
PREVATT
Number of previous auction attempts
TURNOUT
Number of turnout
TERR
Terrace house
SEMID
Semi-detached house
DETACH
Detached house
TEN
Tenure: Freehold (> 99 years) = 1; or leasehold (≤ 99 years)
No bid received = 1
NOBID
SOM
INCREM
State of the market = 1 if real estate market did not
experience two consecutive quarters of negative price
change
Number of increments
VP
property with vacant possession = 1
TPOB
% difference between subsequent sale price and opening bid
TPSB
% difference between subsequent sale price and last bid
TOS
Number of days between auction and subsequent sale
D96, D97, D98, Year dummy variables for 1996 through 2000
D99, D00
KF, JLL, CJ, Auction house dummy variables
DTZ
SUBOUT
subsequent sale = 1
16
Exhibit 1b: Descriptive Statistics
Variable
SUBOUT
LRP
SOM
VP
TEN
TURNOUT
D96
D97
D98
D99
D00
PREVATT
INCREM
TOS
TPOB
TPSB
KF
JLL
CJ
DTZ
TERR
SEMID
DETACH
DISTR
NOBID
LOSS
LOSSOWN
Mean
0.4720
0.0761
0.5414
0.8872
0.8271
183.19
0.0725
0.0393
0.4296
0.3996
0.0104
0.2660
0.3602
138.46
-0.0201
0.0017
0.3530
0.3012
0.0921
0.2226
0.1480
0.2215
0.1801
0.4689
0.7464
0.1108
0.0591
Std Dev
0.4995
0.3207
0.4985
0.3166
0.3783
88.85
0.2594
0.1945
0.4953
0.4901
0.1013
0.6362
1.2328
130.07
0.1570
0.0755
0.4782
0.4590
0.2894
0.4162
0.3553
0.4155
0.3845
0.4993
0.4353
0.1777
0.1404
Min
0
0
0
0
0
15
0
0
0
0
0
0
0
0
-0.5
-0.4632
0
0
0
0
0
0
0
0
0
0
0
Max
1
0.9171
1
1
1
450
1
1
1
1
1
5
12
701
0.7365
0.7094
1
1
1
1
1
1
1
1
1
0.9
0.9
nob
966
966
966
966
966
966
966
966
966
966
966
966
966
456
456
456
966
966
966
966
966
966
966
966
966
583
583
Note:
LOSS = max(original purchase price – estimated value, 0)
LOSSOWN = LOSS*OWN
Where OWN is a dummy variable (=1 if the sale is put up by owners)
17
Exhibit 2a: Analysis of Time-to-Sale and Price-Bid Differences
At least 1 Bid
mean
std dev
111.04
123.04
0.0322
0.1643
0.0065
0.1463
TOS
TPOB
TPSB
nob
122
No Bid
mean std dev
148.48 131.29
-0.0392 0.1500
nob
334
t-stat
-2.7401*
4.3799*
Exhibit 2b: Analysis of Price-Last Bid Differentials by Time-to-Sale
TOS
< 60 days
60 – 90
90 – 180
180 – 360
> 360
At least 1 Bid
mean
std dev
0.0406
0.0938
0.0927
0.1749
0.0508
0.2238
-0.0036
0.1850
-0.1505
0.2186
nob
60
17
23
17
7
mean
-0.0244
-0.0558
-0.0458
-0.0379
-0.1104
No Bid
std dev
0.0800
0.0885
0.1857
0.1641
0.2333
nob
105
42
90
68
32
t-stat
4.7132
4.3336
1.6654
0.7515
-0.4165
Exhibit 2c: Analysis of Price-Last Bid Differentials by Time-to-Sale
TOS
< 60 days
60 – 90
90 – 180
180 – 360
> 360
At least 1 Bid
mean
std dev
0.0128
0.0877
0.0330
0.0618
0.0437
0.2230
-0.0186
0.1713
-0.1601
0.2130
nob
60
17
23
17
7
18
Exhibit 3: Distribution of TPOB by TOS
0.1500
0.1000
0.0500
0.0000
< 60
60 – 90
90 – 180
180 – 360
> 360
-0.0500
-0.1000
-0.1500
-0.2000
1 or more bid
no bid
Exhibit 4: Distribution of TPSB by TOS
0.1000
0.0500
0.0000
< 60
60 – 90
90 – 180
180 – 360
> 360
-0.0500
-0.1000
-0.1500
-0.2000
19
Exhibit 5: Probit Model of Subsequent Sale
Variable
Constant
LRP
SOM
KF
JLL
CJ
DTZ
DISTR
TEN
TURNOUT
D97
D98
D99
D00
TERR
SEMID
DETACH
VP
INCREM
PREVATT
NOBID
LOSS
LOSSOWN
Dependent
Variable
Log-likelihood
nob
Coefficient
0.4850
0.0574
-0.2339
-0.4852
-0.3936
-0.6579
-0.1510
0.0463
0.1258
0.0014
-0.2690
-0.2501
-0.7194
-1.5272
0.2537
0.1537
-0.0756
0.1020
-0.0156
-0.0685
-0.2063
+
*
*
**
**
+
+
(1) Full Sample
Std Error
t-stat
0.3925
1.2360
0.1377
0.4170
0.2721
-0.8600
0.2692
-1.8020
0.2730
-1.4420
0.2971
-2.2140
0.2697
-0.5600
0.0992
0.4670
0.1201
1.0480
0.0006
2.3560
0.2406
-1.1180
0.2723
-0.9180
0.1622
-4.4350
0.5824
-2.6220
0.1323
1.9180
0.1157
1.3290
0.1245
-0.6070
0.1397
0.7300
0.0397
-0.3940
0.0693
-0.9890
0.1129
-1.8280
p-value
0.2166
0.6769
0.3900
0.0715
0.1493
0.0268
0.5756
0.6404
0.2948
0.0185
0.2634
0.3584
0.0000
0.0087
0.0551
0.1839
0.5437
0.4651
0.6939
0.3227
0.0675
(2) Sub-Sample (w/ previous transaction price)
Coefficient
Std Error
t-stat
p-value
0.9226
0.5239
1.7610
0.0782
0.0760
0.1593
0.4770
0.6334
-0.4551
0.4108
-1.1080
0.2679
-0.9147 **
0.3116
-2.9360
0.0033
-0.8894 **
0.3182
-2.7950
0.0052
-1.1255 **
0.3499
-3.2170
0.0013
-0.6801 *
0.3132
-2.1720
0.0299
-0.1074
0.1524
-0.7040
0.4811
0.0695
0.1505
0.4620
0.6442
0.0019 *
0.0008
2.3610
0.0182
-0.0737
0.3639
-0.2030
0.8394
-0.1243
0.4068
-0.3060
0.7599
-0.3410
0.2301
-1.4820
0.1385
-1.1270 **
0.6418
-1.7560
0.0791
0.4779 **
0.1774
2.6930
0.0071
0.1975
0.1607
1.2290
0.2189
-0.1121
0.1674
-0.6700
0.5030
-0.0995
0.1754
-0.5670
0.5707
0.0085
0.0527
0.1620
0.8713
0.0215
0.0821
0.2620
0.7933
-0.2452 +
0.1524
-1.6090
0.1076
0.3298
0.4382
0.7520
0.4518
-0.9545
0.6285
-1.5190
0.1288
SUBOUT
-607.63
966
** significant at 1%
* significant at 5%
+ significant at 10%
-356.03
583
18
Exhibit 6: OLS regression on Price-Open Bid Differential (dependent variable)
(1) Subsequent Sale
(2) Full Sample
Variable
Coeff
Std Error
t-stat
p-value
Coeff
Std Error
t-stat
Constant
0.1336
0.1379
0.9690
0.3332
0.0076
0.0306
0.2490
LRP
-0.0115
0.0130
-0.8840
0.3773
-0.0001
0.0058
-0.0100
SOM
0.0298
0.0487
0.6130
0.5402
0.0062
0.0264
0.2360
KF
0.1083
0.0916
1.1820
0.2377
0.0065
0.0134
0.4880
JLL
0.0754
0.0772
0.9770
0.3289
-0.0006
0.0132
-0.0470
CJ
0.1594
0.1246
1.2790
0.2016
0.0095
0.0146
0.6530
DTZ
0.0579
0.0406
1.4270
0.1544
0.0094
0.0138
0.6850
DISTR
0.0251
0.0175
1.4370
0.1514
0.0113
0.0074
1.5180
TEN
-0.0190
0.0310
-0.6130
0.5399
0.0018
0.0072
0.2520
TURNOUT
0.0000
0.0002
-0.0620
0.9505
0.0001 **
0.0000
2.6730
D96
-0.0477 +
0.0245
-1.9490
0.0519
-0.0249 +
0.0149
-1.6650
D97
-0.0670
0.0587
-1.1400
0.2548
-0.0515 *
0.0218
-2.3600
D98
-0.0342
0.0607
-0.5640
0.5730
-0.0314
0.0298
-1.0560
D99
0.0955
0.1312
0.7280
0.4669
-0.0189
0.0118
-1.6040
D00
0.2684
0.2939
0.9130
0.3616
-0.0265 +
0.0140
-1.8950
TERR
-0.0521
0.0467
-1.1160
0.2650
-0.0001
0.0085
-0.0110
SEMID
-0.0782 **
0.0290
-2.6910
0.0074
-0.0271 **
0.0079
-3.4180
DETACH
-0.0287
0.0236
-1.2190
0.2235
-0.0202 *
0.0089
-2.2720
VP
-0.0055
0.0250
-0.2220
0.8245
-0.0008
0.0097
-0.0820
INCREM
0.0134 *
0.0055
2.4330
0.0154
0.0052 *
0.0026
2.0090
PREVATT
0.0021
0.0155
0.1330
0.8940
-0.0017
0.0041
-0.4190
NOBID
0.0068
0.0386
0.1750
0.8610
-0.0136 +
0.0079
-1.7160
TOS
-0.0001 **
0.0000
-3.9670
0.0001
-0.0002 **
0.0000
-4.7410
Mills Ratio
-0.2663
0.2675
-0.9950
0.3201
-0.0021
0.0040
-0.5270
R-square
0.1925
0.1538
nob
462
972
** significant at 1%
* significant at 5%
+ significant at 10%
All standard errors are heteroskedasticity robust; Mills Ratio is obtained from probit model in Exhibit 5.
p-value
0.8032
0.9922
0.8134
0.6256
0.9625
0.5137
0.4934
0.1291
0.8014
0.0075
0.0959
0.0183
0.2912
0.1088
0.0581
0.9909
0.0006
0.0231
0.9343
0.0446
0.6756
0.0862
0.0000
0.5980
19
Exhibit 7: OLS Regression on Log(Opening Bid) (dependent variable)
Variable
Constant
SOM
VP
TEN
D96
D97
D98
D99
D00
TERR
SEMID
DETACH
PREVATT
RPIA
DISTR
LOSS
LOSSOWN
Coefficient
12.8070
0.1085
-0.2461
0.3293
0.4033
0.4567
0.1788
0.4734
0.3638
0.2101
0.5124
1.2532
-0.0185
0.0026
-0.1155
0.2712
-0.2316
**
**
**
**
**
**
+
**
**
**
**
*
**
Std Error
0.2869
0.1512
0.0618
0.0500
0.1375
0.1655
0.1991
0.1627
0.2033
0.0574
0.0507
0.0616
0.0258
0.0006
0.0461
0.1307
0.2089
t-stat
44.6350
0.7170
-3.9850
6.5850
2.9340
2.7590
0.8980
2.9100
1.7900
3.6600
10.1020
20.3370
-0.7170
4.0300
-2.5050
2.0760
-1.1090
p-value
0.0000
0.4733
0.0001
0.0000
0.0033
0.0058
0.3692
0.0036
0.0735
0.0003
0.0000
0.0000
0.4733
0.0001
0.0122
0.0379
0.2676
R-square
0.6384
nob
583
** significant at 1%
* significant at 5%
+ significant at 10%
All standard errors are heteroskedasticity robust. RPIA is the property price index at time of auction.
18