Journal of Economic Literature
Vol. XLI (September 2003) pp. 763–786
Auctions and the Price of Art
ORLEY ASHENFELTER and KATHRYN GRADDY1
763
1. Introduction
The value of most important works of art
is established by public auction, either
directly, by an actual sale, or indirectly, by
reference to other sales. How the auction
system works is thus a critical determinant of
how the public’s preferences are translated
into the evaluation of artistic work. The auction
system is central in the determination
of the incentives for artistic work, and the
efficiency of the auction system is a key
determinant of the cost of creating and
distributing works of art.
This paper contains a review of the burgeoning
new research of the last decade that
has been designed to shed light on how the
art auction system actually works, what it
indicates about price formation, and how
well it performs. We begin our discussion in
section 2 with a short description of the
mechanics of the auction system and then
organize the remainder of our discussion
around two major topics. The first topic concerns
how auction prices can be used to
determine and compare overall price movements
within the art market and other markets,
and the second topic concerns how the
auction mechanism influences prices. We
1 Ashenfelter: Princeton University and NBER.
Graddy: University of Oxford and CEPR. Our thanks to
John McMillan and the referees for helpful comments,
and also to Alan Krueger and Victor Ginsburgh for comments
on a very early draft of this paper.
address both topics because a deeper understanding
of how prices are affected by the
auction mechanism is important for both
economists and noneconomists seeking to
understand these markets.
We begin section 3 on price-movement
comparisons with a discussion of how art
asset prices can be measured over time. Art
objects are generally unique, so that measuring
time-series movement in their prices
requires careful thought and extensive data.
Once time-series movement is accurately
determined, the data can then be used for a
variety of purposes. A primary goal of the
measurement of time-series movement in
art prices is to evaluate the benefits of
including art assets in a balanced investment
portfolio, and we review the key findings in
section 3.1. We find that in recent years
returns on art assets appear to be little
different from returns on other assets.
Some researchers have recently found that,
because of the weak correlation between
art asset returns with other returns, there
may be a case for the inclusion of art assets
in a diversified portfolio. Since the key
parameters for making this decision are
difficult to estimate, this issue deserves far
more research.
A second use of time-series movement
involves potential pricing anomalies in art
market pricing. The evidence clearly suggests
that, contrary to the view of the art
764 Journal of Economic Literature, Vol. XLI (September 2003)
trade, “masterpieces” often underperform
the market, although the precise interpretation
of this finding is still open for study. In
addition, there is considerable evidence that
there are fairly long periods in which art
prices may diverge across geographic areas
and even auction houses. Finally, we address
a controversial issue regarding price formation
and study whether paintings that are
“bought in” (i.e., those that do not meet their
reserve price) at auction are “burned” (lose
value in subsequent auctions).
In section 4, we turn our attention to the
way the auction mechanism works. We
begin by demonstrating how a misunderstanding
of the auction mechanism resulted
in a curious distribution of the proceeds
from the settlement of the civil suit in the
Christie’s-Sotheby’s price-fixing case. We
show that, contrary to the way the proceeds
from the settlement of the civil suit were distributed,
buyers were almost certainly not
injured by the collusion, but sellers were.
We then review the extensive research on
the effects of the auction institution on price
formation. There is now considerable theoretical
research on strategic behavior in auctions,
much of it in response to empirical
findings, and we review four key findings.
First, the evidence suggests that art experts
provide extremely accurate predictions of
market prices, but these predictions do not
optimally process the publicly available
information. Second, high reserve prices
and the resulting high unsold (“buy-in”)
rates are best explained as optimal search in
the face of stochastic demand. Third, the use
of secret reserve prices remains a puzzle.
Fourth, extensive research has documented
that the prices of identical objects are more
likely to decline than to increase when multiple
units are sold, and this has led to considerable
theoretical research. Subsequent
empirical research has tended to document
declining prices even when the objects are
imperfect substitutes, although the empirical
analysis required in this case must be
more sophisticated.
In section 5, we conclude our review of
studies on art auctions. Because of the
unique nature of many art objects and the
effect of the auction mechanism on price,
the interpretation of market prices requires
great care. Nevertheless, art auctions provide
key information for the evaluation of
artistic work, and they also provide a key laboratory
for testing and refining economic
models of strategic behavior.
2. The Mechanics of Art Auctions
Historically, the major auctioneers of art
have been the English houses of Sotheby’s
and Christie’s. What is called an English auction
is, in fact, Roman. The word auction
comes from the Latin “auctio,” which means
to ascend. The English auction houses
(Christie’s was founded in 1766) have practiced,
refined, and developed many of the
details of the modern auction protocol—and
today they are the dominant forces in the
auction market for art. Almost all art is auctioned
in this ascending price format.
Bidding starts low, and the auctioneer subsequently
calls out higher and higher prices.
When the bidding stops, the item is said to
be “knocked down” or “hammered down,”
and the final price is the “hammer price.”
Not all items that have been put up for
sale and “knocked down” have been sold.
Sellers of individual items will set a reserve
price, which is usually secret, and if the bidding
does not reach this level, the items will
go unsold. Auctioneers say that an unsold
item has been “bought in.” As we show
below, sale rates vary tremendously across
time and across different types of auctions.
An item that has not been sold is rarely, if
ever, actually bought by the auction house. It
may be put up for sale at a later auction, sold
elsewhere, or taken off the market. It is a
part of the auctioneer’s art to “get the bidding
started,” and this may involve accepting
fictitious bids (“off the chandelier” or “from
the order book”) so long as the bidding has
not exceeded the reserve price. Legally, the
Ashenfelter and Graddy: Auctions and the Price of Art 765
auctioneer is bidding on behalf of the seller
when this occurs, but must refrain from
accepting further bids on behalf of the seller
once the bidding exceeds the reserve price.2
Auction houses differ with respect to
whether they announce during the sale
whether an item has been “sold” or is merely
“knocked down” and is unsold. In New York,
all the auction houses have been compelled
by law since the early 1980s to announce
whether the bidding has resulted in a sale.
The practice elsewhere varies by location
and auction house, but there has clearly been
a slow movement toward adopting the practice
originally enacted by law in New York.
While difficult, it is sometimes possible during
an auction, if one listens carefully, to
determine whether an item has been sold or
“bought in.”
Prior to an auction, it is common for a presale
catalogue to be published with information
on the individual items coming up for
sale.3 Included in the pre-sale catalogue is
information on the title of a painting, the
artist, the size of the painting, and the medium.
The auction houses also publish lowand
high-price estimates for the work. The
auction house does not publish, and indeed
is very secretive about, the seller’s reserve
price for the work of art. The auction houses
do commonly observe an unwritten rule of
setting the secret reserve price at or below
the low estimate, but the auctioneer is careful
about revealing anything about the
reserve price during the bidding process.
Auction houses earn income primarily
from commissions charged to buyers and
sellers. The commission charged to buyers is
2 The contractual bases for these arrangements in the
United States are a part of the Uniform Commercial Code,
a set of rules which many states adopt all or part thereof for
purposes of the law of commerce.
3 The auction houses charge a moderate fee for the catalogue,
and participation in the auction without buying the
catalogue would be quite difficult. For less expensive items
that are auctioned, this fee could be thought of as an
entrance fee. However, this interpretation is more difficult
for art, as the cost of the catalogue is truly insignificant
compared to the cost of the items being auctioned.
called the “buyer’s premium.” The total sale
price to the buyer is thus the sum of the
hammer price and the buyer’s premium. In
recent years, published buyer’s premiums
have typically hovered around 10 percent to
17.5 percent of the hammer price of an
object, but they are often lower for large
purchasers. Although buyers may attempt to
negotiate special arrangements regarding
buyer’s premiums, it is our impression that
the typical buyer purchases such a small
fraction of the objects on sale at a particular
auction house that special terms for buyers
are unusual.
Sellers also pay a commission to the auction
house called the “seller’s commission.”
Although the seller’s commission is often
stated as a percentage of the hammer price
(typically 10 percent), it is our impression
that actual seller’s commissions are often
negotiated arrangements that differ by seller.
In some cases, sellers pay no commission
and may even be guaranteed a minimum sale
price. Some key issues related to the negotiation
of seller’s commissions and the extent
of competition and collusion in the setting of
commission rates have recently surfaced in
the trial of Alfred Taubman, former chairman
of Sotheby’s, who was convicted of price
fixing. We discuss issues related to the
Christie’s-Sotheby’s price-fixing case in section
4. We now turn to our discussion of how
prices from art auctions can be used to
construct indices.
3. Art Prices
A key feature of art auctions is that the
items on sale are typically unique, or nearly
so. The result is that there will be some
ambiguity in the construction of a single
index of the movement of prices over time.
One concern about simply using average
prices is that price rises may be exacerbated
during booms as “better” paintings may
come up for sale. For example, Wynne
Kramarsky, whose family formerly owned
Van Gogh’s “Portrait of Dr. Gachet,” said of
766 Journal of Economic Literature, Vol. XLI (September 2003)
the London market prior to the poor sale of
May 15, 1990: “I did not think that London
was poor in terms of performance; I thought
that the pictures were not up to it” (Peter
Watson 1992, p. 10). In general, average art
prices will indicate some variability over
time that is better described as movements
in the heterogeneity of the quality of the
objects offered, rather than as movements in
prices for the same objects.
3.1 Art Price Indices
The extent of heterogeneity, and thus the
ambiguity in the construction of auction
price indices, differs across the items typically
offered for sale by auction. Identical
prints may be offered for sale monthly, while
a given impressionist painting, such as the
“Portrait of Dr. Gachet,” may not be offered
at all in a single decade.
Most art auction indices are based on a
model where the price of the ith
object sold
in time period t is
pit 5 pi 1 pt 1 it,
where pi is the fixed component of the price
that reflects the unique and fixed character
(or “quality”) of the object, pt reflects the
index of aggregate movements in prices, and
the remainder is an idiosyncratic error term.
The key distinction in the construction of
price indices is whether the fixed component
is treated as determined by a small number
of hedonic characteristics, x, that may be
controlled by regression, or whether it is
treated as a parameter that must be
controlled explicitly.
Hedonic models control for the fixed effect
pi with the assumption that pi 5 bxi 1 i,
where i is an error term independent of the
pt’s, and estimate
pit 5 bxi 1 pt 1 i 1 it.
Alternatively, repeat-sale models include a
dummy variable for each painting.
The great attraction of hedonic models is
that all the data may be used in the estimation,
including data on objects that are only
offered for sale once in the sample period.
The disadvantage of these models is the
strong assumption that a (typically small) set
of x variables captures much of the variability
in the fixed components of price (important
if the estimates of the time effects are to
be precise) and that the characteristics of the
objects offered do not vary systematically
over time (important for unbiased estimates
of the time effects). Although the repeat-sale
method overcomes the primary disadvantages
of the hedonic model, it does so at the
cost of discarding data. There must be at
least two observations on a painting’s price,
or it provides no information to help identify
the time index. Indeed, depending on
the frequency with which repeat sales occur,
it may not be possible to identify all the
time effects in the model.
Olivier Chanel, Louis-Andre GerardVaret,
and Victor Ginsburgh (1996) compare
results from repeat-sale and hedonic models.
The overall results indicate that both hedonic
and repeat-sales regressions yield estimates
of real rates of return in art assets over
long intervals that are the same magnitude.
Hence, in some cases the hedonic model
may provide adequate estimates of timeseries
movements in aggregate prices.
However, the danger remains that systematic
movements in the unobserved characteristics
of the objects being offered for sale
may bias the results.
The nature of possible systematic movements
is made clear when we do a detailed
comparison using our data on impressionist
and modern art.4 When yearly price indices
are constructed, the two types of indices at
first appearance are similar. Figure 1 presents
a graph of the hedonic and repeat-sales
price indices for impressionist and modern
art from 1980 to 1991. The correlation
between the two estimates is .9559, the standard
deviation of the hedonic price index is
1.024, and the standard deviation of the
4 See the data appendix for a full description of the
data and the exact way in which the price indices were
constructed.
Ashenfelter and Graddy: Auctions and the Price of Art 767
Figure 1. Repeat-Sale and Hedonic Indices for Impressionist Art
year
repeat-sale price index hedonic price index
1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991
.5
1
1.5
2
2.5
3
3.5
4
repeat-sales index is 1.166. However,
because of movements in the last year, the
two indices give very different internal rates
of return. The hedonic index gives a real
return of about 4 percent, while the repeatsales
price index results in a real return of
about 9 percent! Which is correct? For 1991,
our data ends in May. The “major” impressionist
sales are generally held in October.
One explanation is that the hedonic index
has underestimated the returns for this short
period of time, because it was unable to correct
for quality differences that occur during
sales in the early part of the year. An alternative
explanation is that, because the
repeat-sales index is based on such a small
number of paintings during that period,
these paintings were unrepresentative (i.e.,
their price held up better in poor market
conditions) of the market as a whole. Our
hedonic model incorporates as many as 8792
observations, while the repeat-sale estimates
are based on only 474 observations.
This criticism is not applicable to the sales
of all types of art. For example, the number
of observations that are discarded when
using repeat sales is smaller when studying
the market for prints, as many prints are virtually
identical. James Pesando (1993)
excludes less than one percent of the realized
prices on modern prints when using the
repeat-sale methodology to construct the
print price index.
One can also measure the extent to which
one type of index deviates from the other.
Suppose, for example, that the repeat-sales
index, x, is the true index, x is a measured
hedonic index, and v is independent of x, so
that
x 5 x 1 v .
If x were measured exactly, then a regression
of x on x would give a slope of unity. If x is
measured with error, the difference from
unity provides an estimate of the measurement
error as a fraction of the total variance in
768 Journal of Economic Literature, Vol. XLI (September 2003)
the repeat-sales index. Computing the above
regression, we find that x has a coefficient of
.8400, and a standard error of .086, so the test
for measurement error is on the borderline of
statistical significance. The implication is that
about 16 percent of the variance in the repeatsales
measure of prices is measurement error.
Authors who have calculated price indices
for art include Robert Anderson (1974), John
Stein (1977), William Baumol (1986), Bruno
Frey and Werner Pommerehne (1989), Luc
Buelens and Victor Ginsburgh (1993),
Pesando (1993), William Goetzmann (1993,
1996), Madeleine Barre, Sophie Docclo, and
Victor Ginsburgh (1996), James Pesando and
Pauline Shum (1996), Corinna Czujack
(1997), and Jiangping Mei and Michael
Moses (2001). The details of these studies
and the estimated rates of return on art
assets they contain are presented in table 1.
The estimated returns to holding art are
quite dependent upon the time frame actually
studied, which would be expected. Even
among authors looking at similar time
frames, the returns can vary. The variation
reflects differences in data, and differences
in method. Furthermore, it is difficult to
come to any broad conclusions about the differences
in estimates when using repeatsales
or hedonic indices. Anderson (1974)
finds a real return of 2.6 percent using hedonic
indices and 3.0 percent using repeat
sales on art data from 1780–1960; and
Chanel, Gerard-Veret, and Ginsburgh (1996)
find real returns of 4.9 percent and 5.0 percent
for hedonic and repeat sales indices,
respectively, for the period 1855–1969.
The research reviewed above has focused
on accurate construction of indices that
measure aggregate price movements of
unique objects. As discussed below, these
indices are crucial for answering a variety of
economic questions.
3.2 Art as an Investment
A primary concern of many papers that
construct indices is whether art outperforms
or underperforms stocks and bonds and the
correlation of art investment returns with
other investment portfolios. Once a rate of
return on art assets is calculated based on one
of the price indices above, it is possible to use
this return to decide whether it may be sensible
to include art investments in a diversified
portfolio. Generally, art investments are more
attractive—using the standard capital asset
pricing model (CAPM)—the greater is their
return relative to the return on a risk-free
asset and the weaker the correlation (or beta)
between art investment returns and the
return on other assets. Pesando (1993) used
the standard market model to assess these two
characteristics of art investments in the case of
modern prints. Pesando estimates the model:
Rt
P
2 rf,t 5 a 1 b(Rm,t 2 rf,t) 1 ut
where Rt
P
denotes the return on the print
portfolio, Rm,t denotes the return on the market
portfolio (Pesando uses the S&P 500
stock index), and rf,t denotes the risk-free rate
(Pesando uses 180-day treasury bills).
Pesando estimates a b for the entire print
portfolio of .315, and estimates negative,
but insignificant, risk adjusted returns.
This implies that print investments tend to
reduce the riskiness of a portfolio comprised
of stocks only.
Determining whether art outperforms or
underperforms a market portfolio is not an
easy question to address. First of all, as
Goetzmann (1993) points out, there are many
problems with the calculation of the returns
to art, beginning with selection bias in the
data. As all of the sales prices are drawn from
auction records, only paintings that have been
re-auctioned are included. This excludes both
the high end and the low end of the return
distribution. Paintings that fall drastically in
value or are not generally in demand are generally
not resold at auction; in addition, paintings
that are donated to museums generally
do not reappear. Furthermore, whether or
not an owner decides to sell a painting at auction
may be determined by whether or not
the painting has increased in value. Other
Ashenfelter and Graddy: Auctions and the Price of Art 769
TABLE 1
ESTIMATED RETURNS TO ART FROM VARIOUS STUDIES
Nominal Real
Author Sample Period Method return return
Anderson (1974) paintings in general 1780–1960 hedonic 3.3% 2.6%
paintings in general 1780–1970 repeat sales 3.7% 3.0%
Stein (1977) paintings in general 1946–68 assumes random 10.5%
sampling
Baumol (1986) paintings in general 1652–1961 repeat sales 0.6%
Frey and Pommerehne (1989) paintings in general 1635–1949 repeat sales 1.4%
1950–87 repeat sales 1.7%
Buelens and Ginsburgh (1993) paintings in general 1700–1961 hedonic 0.9%
Pesando (1993) modern prints 1977–91 repeat sales 1.5%
Goetzmann (1993) paintings in general 1716–1986 repeat sales 3.2% 2.0%
Barre et al. (1996) great impressionist 1962–91 hedonic 12.0% 5%
other impressionist 1962–91 hedonic 8.0% 1%
Chanel et al. (1996) paintings in general 1855–1969 hedonic 4.9%
paintings in general 1855–1969 repeat sales 5.0%
Goetzmann (1996) paintings in general 1907–77 repeat sales 5.0%
Pesando and Shum (1996) Picasso prints 1977–93 repeat sales 12.0% 1.4%
Czujack (1997) Picasso paintings 1966–94 hedonic 8.3%
Mei and Moses (2001) American, impressionist, 1875–2000 repeat sales 4.9%
and old masters
Graeser (1993) antique furniture 1967–86 neither 7.0%
Ross and Zondervan (1993) Stradivari violins 1803–1986 hedonic 2.2%
As many of the surveys only report nominal returns, the authors calculated the real return rates as follows. For
the Anderson and Baumol studies, an inflation rate of 0.7 percent a year was used. This number is based on
Baumol’s estimate of inflation during the 300-year period of his study using the Phelps–Brown and Hopkins price
index. Goetzmann’s estimate of inflation during the period of his study (also based on Phelps-Brown and
Hopkins) is 1.2%. French price inflation between 1962 and 1992 according to OECD statistics was 7%.
Assumes random sampling within a portfolio of fixed furniture types.
problems with estimating returns are that
transaction costs are excluded, and in contrast
to stocks and bonds, as we noted above, these
can be quite high (as much as 25 percent of
the value of the object, considering both
buyer’s premiums and seller’s premiums).
Finally, there is significant theft and fire risk
(and hence insurance costs) and cleaning
costs involved in investing in art.
On the other hand, unlike stocks and
bonds, art also pays some dividends in the
form of the pleasure the viewer (and owner)
770 Journal of Economic Literature, Vol. XLI (September 2003)
5 For the CAPM, Mei and Moses follow Campbell
(1987) and estimate
r i,t 1 1 5 Et[ri,t 1 1] 1 k51
K
bik f k,t 1 1 + i,t 1 1,
where r i,t 1 1 is the excess return on asset i held from time
t to time t+1. Et[ri,t 1 1] is the conditional expected return
on asset i, conditional on information known to market
participants at the end of time period t. It is allowed
to vary over time (see Mei and Moses 2001 for details).
f k,t 1 1 are excess returns on k different asset classes.
receives. In principle, the value of these dividends
could be measured by the rental cost of
similar art assets, but we are unaware of any
study that has attempted to do this.
Moreover, it seems unlikely that these returns
would be significant for a large, diversified art
portfolio that is not displayed.
Baumol (1986) and Goetzmann (1993)
tend to concur that art is dominated as an
investment vehicle. Goetzmann writes:
“While returns to art investment have
exceeded inflation for long periods, and
returns in the second half of the twentieth
century have rivalled the stock market, they
are no higher than would be justified by
the extraordinary risks they represent.”
Goetzmann (1993) does not formally estimate
a CAPM, but simply reports correlations
of art returns with inflation, the Bank of
England Rate, consol bond returns, and the
London Stock Exchange.
Although their estimates of the return to
art are not significantly different from previous
estimates, Mei and Moses (2001) take a
different view. They argue that “a diversified
portfolio of artworks may play a somewhat
more important role in portfolio diversification
than discovered in earlier research.”
They base their conclusions on their finding
that their art price index has lower volatility
and a lower correlation with other asset classes
than reported in previous research. They
report that these differences are partly due to
sample selection and partly due to a different
time frame studied. Although Mei and Moses
(2001) estimate a more sophisticated form of
CAPM than has previously been estimated
for art, they primarily base their conclusions
on their estimates of the art index and simple
correlations with bond and stock portfolios.5
Some authors have looked at the financial
returns to holding other collectible items. For
example, Myron Ross and Scott Zondervan
(1989) estimate the real returns to holding
Stradivari violins between 1803 and 1987 to
be 2.2 percent, and Paul Graeser (1993) estimates
returns to holding antique furniture
between 1967 and 1986 to be 7 percent. For
a good survey of papers calculating the rate of
returns in various markets, see Bruno Frey
and Reiner Eichenberger (1995).6
What can we conclude from these studies
on the returns to art? The studies reviewed in
table 1 all report positive returns. However,
returns to art generally appear to be less than
the real rate of return on common stock.
Furthermore, as is clear from figure 1, investing
in art over a short period of time can be
risky.
From current research, however, it is difficult
to conclude whether art should be
included in a diversified portfolio. Many of
the studies listed in table 1 show that the
returns to art may outperform bonds.
Furthermore, the correlation of art to other
investment portfolios may be low. It appears
that different views about the financial benefits
of investments in art assets are primarily
based on empirical issues that revolve, in
part, around the temporal instability and sensitivity
of the estimates of key parameters
related to the market performance of art
investments. This suggests that an important
area for additional research is the development
of a more general empirical model that
will provide an explanation for temporal
instability and thus lead to better-informed
decisions.
3.3 The Masterpiece Effect
Pesando (1993) describes the “masterpiece
effect” by quoting art dealer Edward
6 One other paper of interest is by William Landes
(2000). He calculates the returns to art that two high profile
collectors (Victor and Sally Ganz) achieved. Based on
prices in each of three different auctions that were held
between 1988 and 1997, the real average yearly return
achieved by the Ganz’s was between 12.06 and 21.49.
Ashenfelter and Graddy: Auctions and the Price of Art 771
7 Quote originally taken from Art and Auction
[“Antiques”], Sept. 1988, p. 131.
8 Please see the data appendix for a description of these
datasets and construction of the indices.
Merrin: “. . . it’s always better to buy one
$10,000 object than ten $1,000 objects, or
one $100,000 object—if that is what you can
afford—than ten $10,000 ones.”7 There have
now been several authors who have tested
for the masterpiece effect.
Pesando tests for the effect by constructing
a portfolio of the top 10 or 20 percent of
prints by price, where price is determined
during the first few years of his sample. If
the art trade view is correct, the estimated
price indices for these masterpieces should
uniformly outperform the general portfolio.
He finds no support for this view, and in fact
finds that in part of his sample, masterpieces
provide the lowest cumulative return. Mei
and Moses (2001) find a similar negative
effect for masterpieces, and in fact find this
effect to be uniform across American,
impressionist, and old master samples.
Other authors have found no masterpiece
effect (Goetzmann 1995, and Ginsburgh and
Philippe Jeanfils 1996).
Using our data on impressionist and contemporary
art, we find that it appears that
masterpieces have underperformed in the
contemporary art sample, but not in the
impressionist art sample.8 We construct our
index by dividing each sample into the top
20 percent of paintings sold by price, and the
bottom 80 percent sold by price, and then
construct a hedonic index. Our index is
charted in figures 2 and 3. We find similar
results to Pesando and to Mei and Moses in
contemporary art, but find no effects in
impressionist art. For contemporary art,
masterpieces underperform the lowervalued
paintings by about 5 percent on average
per year, which is quite significant.
There does not appear to be a difference in
the impressionist art dataset.
As pointed out by Pesando (1993), there
should not be a positive masterpiece effect.
An efficient art market should capitalize
favorable properties into their prices, and
their risk-adjusted rates of returns should
not exceed that obtained on other art objects
in the same class. However, the art market
and its participants may not always be efficient.
Werner Pommerehne and Lars Feld
(1997) conclude that museums outside the
United States pay above-average prices in
auction markets. If museums generally purchase
masterpieces and do not resell them,
this could contribute to a positive masterpiece
effect.
Various plausible reasons exist as to why
there should be a negative masterpiece
effect. Mei and Moses (2001) speculate that
a negative masterpiece effect may be due to
overbidding and then mean reversion. This
explanation appears quite reasonable given
the way that various studies above have
defined masterpieces as the highest-priced
paintings that were sold. If a masterpiece is
defined purely by price, there may be some
paintings in the masterpiece sample that
randomly commanded a higher price, perhaps
because two or more bidders had high
private valuations for the paintings. At a later
auction, the prices on these paintings revert
to the mean, thus resulting in a negative
masterpiece effect.
A different explanation for the negative
masterpiece effect may be what Goetzmann
(1996) terms “survivorship bias.” It is likely
that the more expensive paintings remained
in the sample throughout, even if they
decreased in value, whereas less-expensive
paintings dropped out of the sample. Hence
it may appear that masterpieces have underperformed
in the sampled data, but in actuality
less-expensive paintings that have
underperformed are no longer in the sample.
The result that our sample shows a masterpiece
effect for contemporary art but no
masterpiece effect for impressionist art
tends to support the survivorship bias explanation.
As the impressionist art dataset consists
only of “famous” impressionist artists
(defined by those whose paintings are most
often auctioned), it is unlikely that any of
772 Journal of Economic Literature, Vol. XLI (September 2003)
Figure 2. Contemporary Art
year
masterpieces other
1980 1985 1990 1995
.59639
5.77628
Figure 3. Impressionist Art
year
masterpieces other
1980 1985 1990
.77073
2.73552
Ashenfelter and Graddy: Auctions and the Price of Art 773
TABLE 2
THE MASTERPIECE EFFECT
Author Sample Period Result
Pesando (1993) modern prints 1980–92 Return of 11% less
for masterpieces
Ginsburgh and Jeanfils impressionist, modern and contemporary 1962–91 No effect
(1995) European masters, other minor European
painters, contemporary U.S. painters
Goetzmann (1996) paintings in general 1897–1987 No effect
Barre et al. (1996) impressionist 1962–91 Great impressionists
return 4% more than
other impressionists
Mei and Moses (2001) American, impressionist, and 1875–2000 A 10% increase in price
old masters reduces returns by 1%
Ashenfelter and Graddy impressionist art 1980–91 No effect
(see appendix) contemporary art 1982–94 Return of 50% less
for masterpieces
these paintings decreased by so much in
value that they dropped out of the sample.
In contrast, the contemporary art dataset
consists of all paintings that were sold at
auction. It is very likely in this dataset that
some of the paintings in the bottom 80 percent
of prices are no longer in the sample
because they have decreased in value. This
could lead to an upward bias in the estimation
of the “other” sample for contemporary
art, and hence a measured masterpiece
effect.
We have summarized the findings of
papers estimating a masterpiece effect in
table 2. Out of the six studies reported, only
one study, Barre et. al. (1996), finds a positive
masterpiece effect. Interestingly, they
constructed their subsamples slightly differently
than most of the other studies reported,
as they chose masterpieces by reputation of
painter, rather than by price. Nonetheless,
from current research, we can conclude that
if a masterpiece effect exists, it is likely to be
negative rather than positive.
3.4 Is There Evidence that Paintings
Are “Burned?”
As Ashenfelter (1989) noted, it is often
claimed that when an advertised item goes
unsold at auction, its future value will be
affected. Such items are said to have been
“burned.” There has been surprisingly little
work testing this proposition.
Using our repeat-sales data on impressionist
art, we present summary statistics of artworks
that appear twice in the data. We have
looked at whether there appear to be any differences
in prices and estimates for paintings
that came to auction and did not sell during
their first appearance at auction, but sold
during their second appearance at auction
(unsold-sold sample), vs. those that came to
auction, sold during their first appearance,
and were resold again during their second
appearance at auction. We do this by comparing
ratios in the two samples. The ratios
we look at are the estimate during a painting’s
second appearance at auction over the estimate
of the same painting during its first
774 Journal of Economic Literature, Vol. XLI (September 2003)
TABLE 3
REPEAT SALES OF SOLD VS. UNSOLD PAINTINGS
Estimate 2 / Sale Price 2 / Days between Estimate 1 No. in
Estimate 1 Estimate 1 Sales Sample
Unsold-Sold Sample 1.41 1.75 768 83673 178
(2.13) (2.51) (642) (131526)
Sold-Sold Sample 4.71 3.77 1167 201224 231
(12.84) (9.85) (791) (762936)
t-statistic 3.14 3.01 5.33 2.01
(comparing
unsold-sold sample
with sold-sold sample)
(standard deviations in parentheses)
appearance at auction (estimate 2/estimate 1)
and the sale price during the painting’s second
appearance at auction over the estimate
of the same painting during its first appearance
at auction (sale price 2/estimate 1). We
average these over the unsold-sold sample
and the sold-sold sample. We do not correct
for the level of the art price index when the
paintings came to market; hence, any results
are only suggestive.
As reported in table 3, we find a significant
difference between the unsold-sold sample
and the sold-sold sample. We find that the
sale price in the second sale is on average
1.75 the estimate in the first sale, if the painting
was unsold in the first sale, and the sale
price in the second sale is on average 3.77 the
estimate in the first sale, if the painting was
sold in the first sale. Furthermore, it takes
longer for paintings to reappear in the sample
if they were sold the first time around than if
they went unsold the first time around.
These summary statistics suggest that the
future value of a painting that goes unsold at
auction is negatively influenced. More studies
confirming this effect and estimating the
magnitude of this effect would be useful.
3.5 The Law of One Price
The “law of one price” dictates that in the
absence of different transactions costs, no
systematic price differences should exist
between geographically distinct markets.
Several authors have tested for price differences
in different auction houses and in different
geographical locations and have
found that the law of one price does not
always hold.
Pesando (1993) focused on the sale of
identical prints in different markets that
occur within thirty days of each other for the
period 1977–92. For the entire period, he
found that prices were 7 percent higher in
New York than in London, and 10 percent
higher in New York than in Europe.
However, the difference was not statistically
significant for the period 1977–89, while it
was statistically significant at 11 percent and
17 percent in comparisons of New York and
London and New York and Europe, respectively,
between 1989 and 1992. Pesando
(1993) describes the trade explanation as
being the presence of Japanese buyers in the
New York market during that period, though
one would expect any systematic price differences
to disappear when buyers respond
to incentives. Pesando also finds significant
differences among auction houses. For the
entire period, he found that prices average
14 percent higher in Sotheby’s in New York
than at Christie’s in New York, but there was
no difference in the prices of prints at
Ashenfelter and Graddy: Auctions and the Price of Art 775
TABLE 4
THE LAW OF ONE PRICE
Author Sample Period Result
Ashenfelter wine 1986 Differences in prices reflect differences in commission
(1989) rates.
Pesando (1993) modern prints 1977–92 Prices average 14% higher at Sotheby’s NY than at
Christie’s NY. Prices were 7% higher in New York
than in London; prices were 10% higher in New York
than in Europe.
Pesando and Picasso prints 1977–93 Prices average 7% higher at Sotheby’s NY than at
Shum (1996) Christie’s NY. No significant differences in price between
NY and London. Prices were 2% higher in New York
than in London.
Mei and Moses American, impressionist, 1875–2000 Mixed evidence on differences between Sotheby’s and
(2001) and old masters Christie’s. Differences when they do exist are small.
Sotheby’s and Christie’s in London. Mei and
Moses (2001) find mixed evidence on the
law of one price. When they do find price
differences, these differences tend to be
small.
Ashenfelter (1989) studied differences in
prices for wine between auction houses that
were attributed to changes in buyers’ premiums.
In the spring of 1986, buyers’ premiums
were 10 percent at Sotheby’s in
London (and at other locations), but
Christie’s in London had no buyer’s premium.
In the spring of 1986, hammer prices
(that is, not including the buyer’s premium)
at Sotheby’s in London were 12 percent less
than prices at Christie’s in London, likely
reflecting the difference in buyers’ commissions.
In the fall of 1986, Christie’s instituted
a 10-percent buyer’s premium in the
London auctions. In auctions held in the
fall of 1986, there was no difference in
prices, while in an auction held in the
spring of 1987, prices at Sotheby’s in
London were 5 percent higher than at
Christie’s, and in the fall of 1987, prices at
Sotheby’s in London were 4 percent lower
than at Christie’s. In this situation, where
one house imposes a buyer’s premium and
the other house does not, the results clearly
indicate that the incidence of the buyer’s
premium does tend to fall on the sellers! As
described below, significant changes to sellers’
commissions and buyers’ premiums
occurred during the 1990s. These changes
may provide an interesting subject for study
by economists.
A summary of papers testing for the law
of one price is presented in table 4.
The papers reported above test for price
differences between identical paintings sold
at different times at different locations and
houses, nearly identical prints sold during
the same time period at different locations
and houses, or identically described bottles
of wine sold during similar time periods at
different houses. Barre, Docclo, and
Ginsburgh (1996) approach the question
differently. They use the coefficient on auction
house in a hedonic regression to test for
price differences. As they point out,
because quality is largely unobservable to
the econometrician, it is very difficult in
hedonic regressions to disentangle true
price effects from quality effects.
776 Journal of Economic Literature, Vol. XLI (September 2003)
4. The Auction Mechanism and
Price Formation
In the above section, we have shown how
auction prices can be used to determine and
compare price movements. Ideally, these
prices reflect the “true” value of a work of
art, and are independent of the price discovery
mechanism. As discussed below,
however, the auction institution itself, with
commissions, experts, pre-sale estimates,
reserve prices, and sequential sales, can have
a profound influence on the price of art. We
begin this section with a discussion of how a
misunderstanding of the effect of commissions
on price resulted in a curious settlement
of a civil suit. We then proceed to discuss
other aspects of the auction mechanism
in art and how it affects price.
4.1 The Christie’s-Sotheby’s
Price-Fixing Case
Prior to 1995, Sotheby’s and Christie’s
were in fierce competition for consignments
from sellers. At times, they would drastically
cut commission rates paid by sellers (in
many cases to nothing), make donations to
sellers’ favorite charities, and even extend
financial guarantees to the sellers. In March
1995, this competition abruptly ended.
Christie’s announced that it would charge
sellers a fixed nonnegotiable sliding-scale
commission on the sales price, and a month
later Sotheby’s announced the same policies.
Detailed documents kept by Christopher
Davidge, Christie’s former chief executive,
show that the abrupt change was due to a
price-fixing conspiracy. Christie’s cooperated
with the Justice Department, and Sotheby’s
pleaded guilty to price-fixing sellers’ commissions
but maintain their innocence with
respect to fixing buyers’ premiums. In
September 2001, a civil suit was settled
where Sotheby’s and Christie’s agreed to
each pay 256 million dollars to the plaintiffs.
This class-action suit comprises anyone who
had bought or sold items through the auction
houses since 1993.
From an economist’s point of view, the
settlement of the civil suit is interesting, but
appears to be misguided. Although Sotheby’s
did not admit to fixing buyers’ premiums in
the criminal settlement of the case, both
Christie’s and Sotheby’s agreed to each pay
$256 million to both buyers and sellers, with
in all probability about two-thirds going to
buyers and one-third going to sellers. This
amount was calculated taking the pricefixing
of buyers’ premiums into account.
According to In Re Auction Houses Antitrust
Litigation (2001), “The proposed plan of
allocation estimated the overcharges to sellers
as 1 percent of the hammer price, and
those for buyers to be 5 percent of the hammer
price up to and including hammer
prices of $50,000, and $2,500 for buyers at
hammer prices exceeding $50,000. The net
settlement fund would be distributed to
class members pro rata based upon each
class member’s overcharges during the relevant
period.”
Even if Sotheby’s and Christie’s admitted
to colluding on buyers’ premiums, the usual
theory of private value auctions implies that,
to first order, buyers deserve no compensation!
The reason is the following. When a
buyer decides to bid in an ascending price
auction, his strategy should be to bid up to
his reservation price, if necessary. The price
that the winning bidder has to pay is essentially
(epsilon above) the reservation price of
the second-highest bidder. When buyers’
commissions are raised, each buyer should
reduce his reservation price by an equivalent
amount, resulting in a reduction in revenue
to the seller by the amount of the buyer’s
commission. Hence, the entire increase in
buyers’ commissions should fall on the sellers.
Thus, the standard model of private
value auctions implies that the entire settlement
arrangement in the civil suit was mis-
guided!
There are several caveats to this argument.
If sellers’ supplies are elastic, some
sellers may not offer their objects for sale
due to the increased commissions. This
Ashenfelter and Graddy: Auctions and the Price of Art 777
9 Our thanks to Victor Ginsburgh, Patrick Legros, and
an anonymous referee for pointing out exceptions to our
argument.
could result in more buyers competing for
the same item, and the increase in the number
of bidders for each item will push up
the price paid by the winning bidder.
Furthermore, in art auctions, the private
value assumption doesn’t strictly apply. If
some bidders are factoring into their value
estimate the likely future market value of the
piece of art, bidders’ values are correlated.
Bidders may increase their reservation
prices if they believe an increase in commissions
will increase market value in the
future. However, it is unlikely that either of
these effects justifies buyers receiving twothirds
of the $512 million. In addition, it can
be argued that the real losers in this case
were the buyers and sellers who did not
manage to transact because of the pricefixing.
Welfare was clearly reduced, as these
transactions were lost.9
Based on the testimony of Diana Brooks
at the criminal trial of Alfred Taubman, previously
chairman of Sotheby’s board, the
buyers and sellers that managed to transact
were fully compensated. Ms. Brooks estimated
that collusion on sellers’ commissions
resulted in higher profits to Sotheby’s of
some $10–15 million per year. Assuming that
Christie’s received the same increased profits
implies that total damages suffered by
sellers would be on the order of $20–30 million
per year. Assuming the conspiracy lasted
five years (approximately the time period
involved) would suggest total damages of
$100–150 million. Since price-fixing damages
are, by statute, tripled, it appears that
the plaintiffs as a group were amply compensated
for the harm they incurred, especially
in view of the fact that they did not
have to proceed to the uncertainty of a trial.
4.2 Role of Estimates and Experts
Before an auction takes place, in their preauction
catalogues, auction house experts
10 Ashenfelter (2000) defines expert opinion as efficient
if it incorporates all of the publicly available information
that is useful in making predictions. He also provides one
example of inefficient expert opinion.
11 As pointed out by an anonymous referee, this does
not rule out that experts may have known that coordination
was taking place between Christie’s and Sotheby’s. After
the agreement, experts at Sotheby’s began advising dealers
and long-standing clients that concessions to the published
sellers’ fees were no longer available and that Christie’s
had adopted a similar policy. These statements indicate
that experts in the various departments knew that both
auction houses had changed their policies with respect to
sellers’ commissions at approximately the same time. It
would have been relatively easy for the experts to infer that
this change was the result of an illegal agreement.
provide a low- and a high-price estimate for
each item. Determining the accuracy of
these estimates raises some important questions
for the study of the role of expert opinion
in economic decisions.10 Of especial
interest is the motivation of the auctioneer in
choosing the high and low estimates. The
theoretical literature stresses that auctioneers
should provide truthful information
about the items being sold.
Ashenfelter’s (1989) results generally
show that auction houses are truthful; the
average of the auctioneer’s high and low estimate
is highly correlated with the price actually
received. Furthermore, John Abowd and
Ashenfelter (1988) find that auctioneers’
price estimates are far better predictors of
prices fetched than hedonic price functions.
The details of the arrangement for price
fixing revealed by Diana Brooks during the
Christie’s-Sotheby’s price-fixing trial provide
further insight into the role of experts at auction
houses. Brooks reported that at one
point her boss, Alfred Taubman, proposed
that the auction houses collude in providing
clients with similar estimates of the value of
their art. Brooks reported that this was
impossible because she could not simply tell
Christie’s departmental experts, who produce
the estimates, to do a dishonest job.11
While the regressions in Alan Beggs and
Kathryn Graddy (1997) generally uphold
these results, they do find systematic underand
over-predictions. For example, they find
that for contemporary art, more recently
778 Journal of Economic Literature, Vol. XLI (September 2003)
12 Bauwens and Ginsburgh (2000) test explicitly for
“efficiency” of auctioneers’ estimates. They find that
experts do not seem to take advantage of all of the information
that is contained in the sales catalogues.
executed artworks are overvalued and longer
and wider paintings are undervalued. For
impressionist and modern art, they find that
wider, signed, and monogrammed paintings
may be underestimated relative to their
value. One explanation for these findings
may simply be that auction houses are unwittingly
overestimating consumer demand
(and hence willingness to pay) for recent
contemporary art, and underestimating consumer
demand for size! Many people in the
trade express surprise at the strong correlations
that many economists have found
between size and price (see Anderson 1974,
and Beggs and Graddy 1997, for examples).
Other authors have also found that ex-ante
valuations cannot be considered unbiased
predictors of market prices, although it is
our impression that biases are not quantitatively
large when they are precisely estimated.
Luc Bauwens and Ginsburgh (2000)
study 1600 lots of English silver sold
between 1976 and 1991 by Christie’s and
Sotheby’s. They find that Christie’s has a tendency
to underestimate systematically, while
Sotheby’s overvalues inexpensive pieces and
undervalues expensive ones. Chanel,
Gerard-Varet, and Stephanie Vincent (1996)
studied jewellery auctions, and found that
experts have an ex-ante valuation that is
lower than the hammer price for all types of
jewels, except for some watches. They speculate
that some strategic undervaluation is
occurring. These results are interesting, in
part because, as Paul Milgrom and Robert
Weber (1982a) show, in general, for auctioneers,
“honesty is the best policy.”
If price estimates are biased, this raises
some interesting questions about the reason
for the bias. One possibility is simply that the
“experts” make systematic errors because
they are not as “efficient” as the linear predictors
they are being tested against.12
Evidence in favor of this hypothesis would
be the finding that observed biases are not
13 In this formulation, we have implicitly assumed that
the multiple of the standard deviation used to generate the
high estimate is the same as the multiple used to generate
the low estimate.
stable and vary from one sample to another
or from one time period to another. Judging
from the results reported above, there is certainly
some evidence to support this view.
Another possibility is that auctioneers
engage in systematic manipulation of the
estimates for strategic purposes. Brooks’ testimony
in the trial of Alfred Taubman provides
some anecdotal evidence on this issue.
Her testimony suggests that, even when the
two leading auctioneers were engaging in
price fixing, they did not attempt to influence
the art appraisers who worked for them
to assist in the conspiracy. However, Mei and
Moses (2002) provide empirical evidence of
possible estimate manipulation. They find
that price estimates tend to have an upward
bias for expensive paintings, and furthermore,
high estimates at the time of purchase
are associated with adverse future abnormal
returns. They interpret these findings as evidence
that auction houses strategically set
estimates and that investors are credulous.
A related question is “what motivates the
auctioneers when they determine the spread
between the high and the low estimates that
are published in the pre-sale catalogues?”
One likely explanation of how the spread is
determined is by the auctioneer’s estimate of
the uncertainty or possible variance in the
price of the painting. In this case, the high
estimate might reasonably be interpreted
as the estimate of the mean price plus a multiple
of the estimated standard deviation
(H5 m 1 rs). Likewise, under this interpretation,
the low estimate would be the mean
minus a multiple of the standard deviation
(L 5 m 2 rs). With this interpretation the
high estimate minus the low estimate divided
by 2 is proportional to the estimated
standard deviation ((H 2 L)/2 5 rs) and the
average of the high estimate and the low
estimate would be the estimated mean
((H 1 L)/2 5 m).13 A large difference in the
Ashenfelter and Graddy: Auctions and the Price of Art 779
14 There can be many explanations for why an art item
fails to meet its reserve price. For example, new information
might have become available during the preview
regarding condition, amount of restoration, or another
attribute that adversely affected market price. Alternatively,
the auction house may have deferred to a consigner’s
desire to place an aggressively high reserve on an item.
Another explanation may be that fewer bidders showed up
than anticipated, due to either random events or a general
economic downturn.
high estimate and the low estimate would
therefore signal a high estimate of price variance
or a lot of uncertainty. However, as the
seller’s secret reserve price, by convention,
lies below the low estimate, it may be that
the spread between the high and low estimate
is not simply a reflection of the auctioneer’s
uncertainty surrounding the possible
price. If the seller wishes to set a high
reserve price, the auctioneer may increase
the low estimate. Ashenfelter, Graddy, and
Margaret Stevens (2002) find some evidence
that this may be occurring. In their datasets
of contemporary art and impressionist and
modern art, a smaller spread between the
high estimate and the low estimate is positively
correlated with an item being bought
in (i.e. the reserve price not being met). This
finding can be interpreted as supporting the
idea that the auctioneer may increase his low
estimate if the seller wishes to set a high
reserve price.
In summary, the evidence on the role of
estimates appears to be divided. Some evidence
indicates that auctioneers are simply
trying to truthfully predict the price of a
painting, while other evidence suggests that
auctioneers may in some cases be altering
the estimate from the true predicted value.
A summary of papers addressing the role of
estimates, along with the varied results, is
presented in table 5.
4.3 Sales Rates and Reserve Prices
As we noted above, items that are put up
for sale at auction often go unsold because
the bidding in the auction does not meet
the reserve price.14 Sale rates vary tremen-
15 The authors obtained these data directly from
Christie’s in London.
dously over time and they also vary systematically
across different types of auctions.
Table 6 shows sale rates in different departments
at Christie’s in London in 1995 and
1996 along with average value of a lot sold.15
As can be seen from the table, 96 percent of
items put up for sale in auctions of arms
and armour were sold, 89 percent of wine at
auction was sold, and 71 percent of impressionist
and modern art items were sold.
Ashenfelter, Graddy, and Stevens (2002)
provide a study of sale rates across time in
art auctions and across different types of
auctions. Based on the observation that an
item is bought in if and only if it does not
meet or exceed its reserve price, they develop
a model of optimal reserve prices. The
seller of a painting faces the following problem:
if he participates in an auction, the
highest bid for the painting can be regarded
as a random draw from some price distribution.
When a seller sets a reserve price, he
must decide at what price he would be indifferent
between selling now and waiting for
the next auction. The optimal policy is to set
a reserve price that is a constant proportion
of the current expected price. Sale rates can
then be modelled as being explained by
price shocks and a constant, or “natural sale
rate.” This natural sale rate (which may vary
across different types of auctions) depends
only on the variance of log prices and the
seller’s discount rate. They estimate that the
reserve price is generally set to be about 70
to 80 percent of the auctioneer’s low estimate.
Although reserve prices are generally
secret, the available evidence suggests that
this prediction is reasonably accurate.
David Genesove (1995) tests a related but
somewhat different theory in the context of
wholesale automobile auctions. He finds
that on average sale rates in used auto auctions
are actually quite low; between about
58 percent and 68 percent of automobiles go
780 Journal of Economic Literature, Vol. XLI (September 2003)
TABLE 5
ROLE OF ESTIMATES
Author Sample Period Result
Milgrom and Weber Honesty is the best policy.
(1982a)
Abowd and Ashenfelter impressionist art 1980–82 Auctioneers’ price estimates are far better
(1988) predictors of prices than hedonic models.
Ashenfelter (1989) impressionist art 1980–82 Auction houses are truthful.
Chanel et al. (1996) jewelry 1993–94 Pre-sale estimates undervalue most types of jewelry,
with the exception of some watches.
Beggs and Graddy (1997) impressionist art 1980–91 Systematic over- and undervaluations (recently
contemporary art 1980–94 executed works of art tend to be overvalued; longer
and wider paintings are undervalued).
Bauwens and Ginsburgh English silver 1976–90 English silver: Christie’s systematically
(2000) underestimates. Sotheby’s overvalues inexpensive
pieces and undervalues expensive pieces.
Ashenfelter et al. (2001) impressionist art 1980–91 Examines whether spread between high and low
contemporary art 1982–94 estimate is indication of auctioneer’s uncertainty
or reflects seller’s wish to set a high reserve price.
Mei and Moses (2002) American, 1950–2002 Auctioneers strategically set estimates in order to
impressionist, and increase realized prices.
old masters
unsold. In his paper, he tests a result by
Ronald Balvers (1990) that also states that an
increase in variance decreases the probability
of sale. He finds that an increase in the
log-variance is associated with a lower probability
of sale, and hence the “natural sale
rate” is again dependent on variance of log
prices.
A related question is whether the reserve
prices that are set in art auctions are optimal.
This question has not been looked at in the
context of art auctions, but R. Preston
McAfee, Daniel Quan, and Daniel Vincent
(2000) derive a lower bound on the optimal
reserve price for a general auction model
with affiliated signals, common components
to valuations, and endogenous entry (all
characteristics which can be applied to art
auctions or other auctions of cultural
objects). They apply their computations to
FDIC real estate auctions and find that the
lower bound on the optimal reserve price for
real estate is about 75 percent of the
appraised value.
Overall, there has been little research into
why sales rates differ between items and
whether reserve prices are optimal. Given
the persistent differences in sales rates
between items (which suggests differing
reserve prices), more research in these areas
would be useful.
4.4 Why Secret Reserve Prices?
An important institutional detail in art
auctions is the use of secret reserve prices.
Auctioneers generally do not reveal the
reserve price, and they make it as difficult
as they can for bidders to infer it. A reserve
Ashenfelter and Graddy: Auctions and the Price of Art 781
TABLE 6
AVERAGE SALE RATES BY DEPARTMENT
No. of
Average Sold Auctions Sale Rate
Department Lot Value in Sample (% of Lots Sold) % Sold by Value
1996 1995 Mean Std. dev Mean Std. dev.
Impressionist £122,820 £135,430 8 71% (0.11) 80% (0.10)
Old masters drawings £50,670 £29,210 4 77% (0.09) 89% (0.08)
Contemporary £36,820 £36,840 7 79% (0.04) 87% (0.06)
British pictures £29,710 £23,560 7 78% (0.14) 83% (0.17)
Old master pictures £29,180 £6,560 11 73% (0.15) 82% (0.15)
Continental pictures £21,810 £10,450 7 72% (0.11) 79% (0.10)
Clocks £14,340 £5,130 4 88% (0.03) 89% (0.07)
Jewellery £12,190 £6,750 8 86% (0.05) 89% (0.04)
Furniture £11,670 £8,220 25 85% (0.09) 92% (0.06)
Silver £11,080 £5,910 10 87% (0.11) 92% (0.07)
Sculpture £11,070 £6,340 5 78% (0.21) 81% (0.20)
Modern British pictures £10,340 £7,190 9 70% (0.05) 81% (0.05)
Victorian pictures £9,460 £8,400 6 66% (0.13) 75% (0.11)
British drawings & watercolors £9,160 £3,400 14 72% (0.14) 87% (0.10)
Rugs & carpets £9,160 £3,700 8 80% (0.17) 85% (0.14)
Topographical pictures £8,640 £8,010 2 68% (0.13) 81% (0.00)
Islamic works £6,670 £6,950 5 68% (0.22) 82% (0.12)
Cars £5,750 £7,610 6 71% (0.16) 65% (0.22)
Chinese works of art £5,640 £6,400 8 70% (0.19) 79% (0.16)
Books & manuscripts £5,220 £4,270 15 81% (0.12) 86% (0.09)
Russian works of art £4,490 £5,480 4 64% (0.14) 69% (0.15)
Japanese works £4,410 £2,840 5 72% (0.04) 76% (0.05)
Musical instruments £3,960 £4,110 5 77% (0.05) 76% (0.16)
Watches £3,870 £2,190 6 71% (0.09) 81% (0.11)
Prints—old modern & contemp. £3,850 £4,230 8 81% (0.12) 92% (0.09)
Miniatures £3,350 £3,260 2 82% (0.05) 92% (0.07)
Antiquities £3,260 £3,640 3 57% (0.08) 66% (0.13)
Porcelain and glass £2,700 £2,600 14 76% (0.12) 85% (0.10)
Tribal art £2,650 £2,090 3 67% (0.08) 75% (0.19)
Photographica £2,580 £1,660 3 61% (0.27) 79% (0.08)
Modern guns £2,510 £3,620 5 93% (0.06) 94% (0.04)
Garden statuary £2,120 £1,540 4 91% (0.10) 91% (0.11)
Arms & armour £1,890 £2,400 4 96% (0.03) 99% (0.01)
Frames £1,800 £2,260 4 81% (0.15) 85% (0.14)
Stamps £830 £650 22 78% (0.13) 82% (0.12)
Wine £690 £580 37 89% (0.09) 91% (0.08)
price clearly contains information about the
seller’s valuation of an item; intuitively,
revealing information matters if the items
contain a common value component among
buyers. While people buy art for enjoyment,
there is an investment component to
many buyers’ motives; that investment
component leads one to classify art as having
common-value components. Thus, the
fact that auctioneers tend to keep reserve
prices secret has been an interesting topic
for discussion since the publication of
Milgrom and Weber’s (1982a) paper, which
showed that it is optimal for a seller of a
good at a common-value auction to reveal
their valuation.
782 Journal of Economic Literature, Vol. XLI (September 2003)
One reason that has been suggested for
secret reserve prices is that these may be
used to deter collusion. As Ashenfelter
(1989) suggested, when the turnout is low,
some sellers may prefer that their goods be
bought in and offered for sale at a later date
rather than risk a collusive ring bidding to
depress the item’s price. If there is a ring
operating, a secret reserve price might
encourage bidders to bid higher than they
would have otherwise.
Vincent (1995) cleverly built upon (and
overturned) the intuition from Milgrom and
Weber’s (1982a) original result. His explanation
is based upon the inhibiting effect that
the announcement of a reserve price may
have on the participation of bidders in a
given auction. This announcement could discourage
some bidders from participating.
As revelation of information is important
for increasing revenues in a common value
auction, the fact that these bidders are not
participating prevents their information
from playing a part in the auction and may
lower overall bids. Hence, there is a tradeoff
between the reserve price revealing the
seller’s information, and a reserve price discouraging
participation, which lowers total
aggregation of information.
4.5 The Declining Price Anomaly
Since Ashenfelter (1989) showed that
prices are twice as likely to decrease as to
increase for identical bottles of wine sold in
same lot sizes at auction, there has been a
tremendous amount of study of the declining
price anomaly in many types of auctions.
The literature on the declining price anomaly
is extensive. Pesando and Shum (1996)
and Beggs and Graddy (1997) address
declining prices in art auctions, and much of
the other literature is applicable to art. We
review this literature below. We begin with a
brief review of the theoretical literature on
why declining prices may occur, and then
review the types of auctions where declining
prices have been found.
Soon after publication of Ashenfelter’s
(1989) article, there were many theoretical
papers written to explain declining prices.
Jane Black and David deMeza (1992)
claimed it was no anomaly; declining prices
in wine auctions exist primarily because the
winner of the first auction in a sequence has
the option to buy the remaining objects at
the winning price. However, this theory is
unable to explain why the anomaly continues
to exist even where this option is not permitted.
McAfee and Vincent (1993) showed that
risk aversion could create declining prices.
One unappealing feature of their explanation
is that a pure-strategy equilibrium exists
only when there is nondecreasing absolute
risk aversion, which is usually thought
implausible. Mixed strategy equilibria are
ex-post inefficient, which is sometimes also
thought to be a weakness of this theory, but
which may nevertheless be a correct characterization
of the actual market. Beggs and
Graddy (1997) attribute a declining price to
pre-sale estimate ratio throughout an auction
to the fact that the value of art auctioned
(as measured by the pre-sale estimate), on
average, declines throughout the auction.
Other theoretical papers are as follows.
Nils-Henrik von der Fehr (1994) shows that
participation costs could create declining
prices through strategic bidding. Richard
Engelbrecht-Wiggans (1994), Dan Bernhardt,
and David Scoones (1994), Ian Gale
and Donald Hausch (1994), and Paul
Pezanis-Christou (2001) relate the price
decline to heterogeneity among buyers, and
Ginsburgh (1998) shows that the presence of
absentee bidders can generate declining
prices.
The declining price anomaly has been
documented in a number of different types
of auctions with different auction structures.
Stephen Buccola (1982) found it occurring
in livestock auctions; Milgrom and Weber
(1982b) for transponder leases; McAfee and
Vincent (1993) and Albert Di Vittorio and
Ginsburgh (1994) confirmed Ashenfelter’s
(1989) wine findings; Stuart Thiel and Glenn
Ashenfelter and Graddy: Auctions and the Price of Art 783
Petry (1990) in stamp auctions; Ashenfelter
and Genesove (1992) and Bruce
Vanderporten (1992a,b) for condominiums;
Engelbrecht-Wiggans and Charles Kahn
(1992) for dairy cattle; Kenneth Lusht
(1994) for commercial real estate; Chanel,
Gerard-Varet, and Vincent (1996) for gold
jewellery; Pesando and Shum (1996) for
Picasso prints; Norman Thurston (1997) for
mink pelts; Pezanis-Christou (2001) for fish
auctions; Gerard van den Berg, Jan van
Ours, and Menno Pradhan (2001) for Dutch
flower auctions; and Ginsburgh and van
Ours (2001) for Chinese porcelain recovered
from shipwrecks. Penny Burns (1985) and
Claudia Keser and Mark Olson (1996) set up
experiments that have reached the same
conclusions.
Several authors have also found increasing
prices. Among them are Neil Gandal (1997)
for Israeli cable television licenses, and
Stephen Donald, Harry Paarsch, and Jacques
Robert (1997) for Siberian timber-export
permits. Chris Jones, Flavio Menezes, and
Francis Vella (1996) found that prices could
increase or decrease in sequential auctions of
wool, as did Chanel, Gerard-Varet, and
Vincent (1996) for watches; Milgrom and
Weber (1982b) show theoretically that if bidders’
valuations are affiliated, then prices will
tend to rise over time in a sequence of auctions
of identical objects. George Deltas and
Georgia Kosmopoulou (2001) find in a sale of
library books that expected prices increase
over the auction, but that probability of sale
decreases. They attribute their findings to
“catalogue” effects: it is important how and
where and item appears in the pre-sale catalogue.
Plamen Natzkoff (2001) provides an
excellent survey of papers on the declining
price anomaly.
It is an interesting result that in a variety
of different types of auctions, price direction
throughout an auction can be predicted. As
reported in table 7, declining prices (on
average) have been documented in more
types of auctions than have rising prices. A
variety of economic theories have been
developed to explain price direction, and in
all likelihood, the price direction results
from a combination of these effects. Declining
prices do not occur in every auction
or every art auction, but they appear to be an
important effect that the auction mechanism
has on price.
5. Conclusion
The empirical study of art auctions really
has two purposes. On the one hand, the auction
mechanism provides a public report on
the prices of art objects. As we have shown,
because of the unique nature of many art
objects, the interpretation of market prices
requires great care. Nevertheless, this information
is the primary way that art objects are
valued and it provides us with our primary
objective information on preferences regarding
art. Although the market is surely
not all that is important in the judgement of
art and artists, it is certainly one of the key
components of our understanding of what is
good and bad.
The empirical study of art auctions also
has another purpose. Art auctions provide
data that may be used to test and refine
strategic models of behavior. Here the
object of study is the economic mechanism,
and it makes very little difference what
object is for sale. It appears that a great deal
of what we know about the operation of
auction mechanisms may also lead to the
rather happy study of objects of considerable
interest in their own right.
The empirical study of art auctions and
the price of art assets has been a growth field
in the last decade and has resulted in an
increasing sophistication in the questions
being asked and in the empirical methods
being used. It seems likely that this trend
will continue into the future.
Data Appendix
The dataset on impressionist and modern art auctions
was constructed by Orley Ashenfelter and
Andrew Richardson. This dataset is restricted to 58
784 Journal of Economic Literature, Vol. XLI (September 2003)
TABLE 7
DECLINING PRICE ANOMALY
Empirical Work (Declining Prices)
Buccola (1982) Livestock
Burns (1985) Experimental results
Ashenfelter (1989) Wine
Milgrom and Weber (1982b) Transponder leases
Thiel and Petry (1990) Stamps
Ashenfelter and Genesove (1992) Condominiums
Venderporten (1992a,b) Condominiums
Engelbrecht-Wiggans and Kahn (1992) Dairy cattle
McAfee and Vincent (1993) Wine
De Vittorio and Ginsburgh (1994) Wine
Lusht (1994) Commercial real estate
Chanel et al. (1996) Gold jewelry
Pesando and Shum (1996) Picasso prints
Keser and Olson (1996) Experimental results
Beggs and Graddy (1997) Art
Thurston (1997) Mink pelts
Pezanis-Christou (2001) Fish
van den Berg et al. (2001) Flowers
Ginsburgh and van Ours (2001) Chinese porcelain from shipwrecks
Empirical Work (Increasing Prices)
Jones et al. (1996) Wool auctions
Chanel et al. (1996) Watches
Gandal (1997) Isreali cable television auctions
Donald et al. (1997) Siberian timber auctions
Deltas and Kosmopoulou (2001) Library books
Theoretical Work
Black and de Meza (1992) Declining prices in wine auctions are due to buyers’ options.
McAfee and Vincent (1993) Risk aversion could create declining prices.
Von der Fehr (1994) Participation costs could create declining prices.
Englebrecht-Wiggins (1994) Relate price decline to heterogeneity of objects.
Bernhardt and Scoones (1994) Relate price decline to heterogeneity of objects.
Gale and Hausch (1994) Relate price decline to heterogeneity of objects.
Beggs and Graddy (1994) Ordering by value can generate price/estimate declines.
Ginsburgh (1998) Absentee bidders can generate declining prices.
selected impressionist and modern artists and includes
only paintings, not sculptures. These artists were chosen
primarily because their work is well represented at
auction. The period covered is 1980 to 1990, and the
dataset includes over 16,000 items in 150 auctions that
were held in London and New York, at both Christie’s
and Sotheby’s. The auction prices were collected from
public price lists, and the estimated prices and observable
painting characteristics were collected from the
pre-sale catalogues. This dataset does not include all
items sold in each auction, only a sample of the 58
selected artists. Furthermore, we only have prices for
items that were sold at auction. This dataset was used
in Richardson (1992), Abowd and Ashenfelter (1989),
Beggs and Graddy (1997), and Ashenfelter, Graddy,
and Stevens (2002).
To construct a repeat-sales price index, we identified
230 paintings that sold at least twice during the period
1980–90 (for a total of 474 observations). To make a
positive identification, we required that paintings have
an identical title, medium, artist and painting date. As
many paintings have identical titles, title and artist are
not sufficient identifiers. We regress the log of the sale
price of the painting on a dummy variable for each
painting, and the time period in which the painting was
sold. We include a dummy variable for each year. Using
the antilogs of the coefficients on the time dummies,
we construct our repeat sales price index for impressionist
art as reported in figure 1.
For the hedonic price index in figure 1, the log of
the sale price is regressed on the hedonic painting
characteristics in addition to time dummies for each
Ashenfelter and Graddy: Auctions and the Price of Art 785
period. The hedonic characteristics used for impressionist
and modern art are painting date, length, width,
signed, monogrammed, stamped, medium in which it
was painted, and artist. We also include dummy variables
indicating whether the painting was sold at
Sotheby’s or Christie’s and New York or London. For
figure 3, the sample is split into the top 20 percent of
paintings sold by price and the bottom 80 percent sold
by price and the index is constructed for these subsamples
as described above.
The dataset on contemporary art was constructed by
Kathryn Graddy and includes all sales of contemporary
art at Christie’s auction house on King Street in London
between 1982 and 1994. The data were gathered from
the archives of Christie’s auction house, and for each
item, the observable characteristics were hand-copied
from the pre-sale catalogues. The information on
whether or not a lot is sold and the final bid from 1988
onwards was taken primarily from Christie’s internal
property system. Before 1988, many of the lots were
missing from the internal system. It appeared that,
after a certain period of time, some of the lots were
deleted from the system, for no predictable reason.
From December 1982 through December 1987, access
to the auctioneer’s books was obtained and used to
track the missing items. The contemporary art dataset
includes 35 auctions and approximately 4500 items for
sale. This dataset was used in Beggs and Graddy (1997)
and Ashenfelter, Graddy, and Stevens (2002).
In order to construct the hedonic indices for contemporary
art that were used in figure 2, the sample is
split into the top 20 percent of paintings sold by price
and the bottom 80 percent sold by price, and the index
is constructed as described above using the following
hedonic characteristics: painting date, length, width,
medium, artist, and whether or not a painting is subject
to VAT. As many artists in the contemporary art sample
have come up for sale just one or a few times, we use a
dummy variable to indicate whether paintings have
been included for sale four or fewer times rather than
using artist dummies for artists with four or fewer sales.
Only the prices of items that were actually sold at auction
were included in the regressions.
REFERENCES
Abowd, John and Orley Ashenfelter. 1988. “Art
Auctions: Prices Indices and Sale Rates for
Impressionist and Contemporary Pictures,” mimeo,
econ. dept., Princeton U.
Anderson, Robert C. 1974. “Paintings as an Investment,”
Econ. Inquiry 12:1, pp. 13–26.
Ashenfelter, Orley. 1989. “How Auctions Work for
Wine and Art,” J. Econ. Perspect. 3:3, pp. 23–36.
———. 2000. “The Demand for Expert Opinion:
Bordeaux Wine,” Les Cahiers de L’OCVE, Cahier
Scientifique 3:Mars.
Ashenfelter, Orley and David Genesove. 1992. “Testing
for Price Anomalies in Real-Estate Auctions,” Amer.
Econ. Rev. 82, pp. 501–505.
Ashenfelter, Orley; Kathryn Graddy and Margaret
Stevens. 2002. “A Study of Sale Rates and Prices in
Impressionist and Contemporary Art Auctions,”
mimeo, econ. dept., U. Oxford.
Balvers, Ronald J. 1990. “Variability and the Duration
of Search,” Int. Econ. Rev. 31:3, pp. 747–51.
de la Barre, Madeleine; Sophie Docclo and Victor
Ginsburgh. 1996. “Returns of Impressionist, Modern
and Contemporary European Paintings 1962–1991,”
Ann. Econ. Statist. 35, pp. 143–81.
Baumol, William J. 1986. “Unnatural Value: Or Art
Investment as Floating Crap Game,” Amer. Econ.
Rev. Papers Proceedings 76:2, pp. 10–14.
Bauwens, Luc and Victor Ginsburgh. 2000. “Art
Experts and Auctions: Are Pre-sale Estimates
Unbiased and Fully Informative?” Rech. Econ.
Louvain 66:2, pp. 131–44.
Beggs, Alan and Kathryn Graddy. 1997. “Declining
Values and the Afternoon Effect: Evidence from Art
Auctions,” Rand J. Econ. 28:3, pp. 544–65.
van den Berg, Gerard; Jan C. van Ours and Menno P.
Pradhan. 2001. “The Declining Price Anomaly in
Dutch Rose Auctions,” Amer. Econ. Rev. 91:4, pp.
1055–62.
Bernhardt, Dan and David Scoones. 1994. “A Note on
Sequential Auctions,” Amer. Econ. Rev. 84:3, pp.
653–57.
Black, Jane and David de Meza. 1992. “Systematic
Price Differences between Successive Auctions Are
No Anomaly,” J. Econ. Manage. Strategy 1:4, pp.
607–28.
Buccola, Stephen. 1982. “Price Trends at Livestock
Auctions,” Amer. J. Agr. Econ. 64:1, pp. 63–69.
Buelens, Nathalie and Victor Ginsburgh. 1993.
“Revisiting Baumol’s ‘Art as Floating Crap Game’,”
Europ. Econ. Rev. 37:7, pp. 1351–71.
Burns, Penny. 1985. “Experience and Decision
Making: A Comparison of Students and Businessmen
in a Simulated Progressive Auction,” in Research
in Experimental Economics: A Research
Annual. Vol. 3. Vernon Smith, ed. Greenwich, CT:
JAI, pp. 139–57.
Campbell, John. 1987. “Stock Returns and the Term
Structure,” J. Finan. Econ. 18:2, pp. 373–99.
Chanel, Olivier; Louis-Andre Gerard-Varet and Victor
Ginsburgh. 1996. “The Relevance of Hedonic Price
Indices,” J. Cult. Econ. 20:1, pp. 1–24.
Chanel, Olivier; Louis-Andre Gérard-Varet and
Stephanie Vincent. 1996. “Auction Theory and
Practice: Evidence from the Market for Jewellery,” in
Economics of the Arts: Selected Essays. Victor
Ginsburgh and Pierre-Michel Menger, eds.
Amsterdam: Elsevier, pp. 135–49.
Czujack, Corinna. 1997. “Picasso Paintings at Auction,
1963–1994,” J. Cult. Econ. 21:3, pp. 229–47.
Deltas, George and Georgia Kosmopoulou. 2001.
“Bidding in Sequential Auctions: ‘Catalogue’ vs.
‘Order of Sale’ Effects (in a Sale of Library Books),”
mimeo, econ. dept., U. Illinois.
Di Vittorio, Albert and Victor Ginsburgh. 1994.
“Pricing Red Wines of Médoc Vintages from 1949 to
1989 at Christie’s Auctions,” mimeo, U. Libre de
Bruxelles.
786 Journal of Economic Literature, Vol. XLI (September 2003)
Donald, Stephen; Harry Paarsch and Jacques Robert.
1997. “Identification, Estimation, and Testing in
Empirical Models of Sequential, Ascending-Price
Auctions with Multi-Unit Demand: An Application
to Siberian Timber-Export Permits,” mimeo, Boston
U., U. Iowa and U. Montréal.
Engelbrecht-Wiggans, Richard. 1994. “Sequential
Auctions of Stochastically Equivalent Objects,” Econ.
Letters 44:1–2, pp. 87–90.
Engelbrecht-Wiggans, Richard and Charles Kahn.
1992. “An Empirical Analysis of Dairy Cattle
Auctions,” mimeo, U. Illinois.
von der Fehr, Nils-Henrik Morch. 1994. “Predatory
Bidding in Sequential Auctions,” Oxford Econ.
Papers 46:3, pp. 345–56.
Frey, Bruno S. and Reiner Eichenberger. 1995. “On
the Rate of Return in the Art Market: Survey and
Evaluation,” Europ. Econ. Rev. 39:3–4, pp. 528–37.
Frey, Bruno S. and Werner W. Pommerehne. 1989.
Muses and Marets, Explorations in the Economics of
the Arts. Oxford: Blackwells.
Gale, Ian and Donald Hausch. 1994. “Bottom-Fishing
and Declining Prices in Sequential Auctions,” Games
Econ. Behav. 7:3, pp. 318–31.
Gandal, Neil. 1997. “Sequential Auctions of Interdependent
Objects: Israeli Cable Television
Licenses,” J. Indust. Econ. 45, pp. 227–44.
Genesove, David. 1995. “Search at Wholesale Auto
Auctions,” Quart. J. Econ. 110:1, pp. 23–49.
Ginsburgh, Victor. 1998. “Absentee Bidders and the
Declining Price Anomaly in Wine Auctions,” J. Polit.
Econ. 106:6, pp. 1302–19.
Ginsburgh, Victor and Philippe Jeanfils. 1995. “Longterm
Co-movements in International Markets for
Paintings,” Europ. Econ. Rev. 39:3–4, pp. 538–48.
Ginsburgh, Victor and Jan van Ours. 2001. “On
Organizing a Sequential Auction: The Case of
Shipwreck Findings,” mimeo, ECARES, U. Libre de
Bruxelles.
Goetzmann, William N. 1993. “Accounting for Taste:
Art and Financial Markets over Three Centuries,”
Amer. Econ. Rev. 83:5, pp. 1370–76.
———. 1996. “How Costly Is the Fall from Fashion?
Survivorship Bias in the Painting Market,”
Economics of the Arts: Selected Essays. Victor
Ginsburgh and Pierre-Michel Menger, eds.
Amsterdam: Elsevier, pp. 71–83.
Graeser, Paul. 1993. “Rate of Return to Investment in
American Antique Furniture,” South. Econ. J. 59:4,
pp. 817–21.
In Re Auction Houses Antitrust Litigation, 2001 U.S.
Dist. LEXIS 1713 (S.D.N.Y)
Jones, Chris; Flavio Menezes and Francis Vella. 1996.
“Auction Price Anomalies: Evidence from Wool
Auctions in Australia,” work. paper 303, Australian
National U.
Keser, Claudia and Mark Olson. 1996. “Experimental
Examination of the Declining price Anomaly,” in
Economics of the Arts: Selected Essays. Victor
Ginsburgh and Pierre-Michel Menger, eds.
Amsterdam: Elsevier, pp. 151–75.
Landes, William. 2000. “Winning the Art Lottery: The
Economic Returns to the Ganz Collection,”
Recherches Economiques de Louvain 66:2, pp.
111–130.
Lusht, Kenneth. 1994. “Order and Price in a
Sequential Auction,” J. Real Estate Finance Econ. 8,
pp. 259–66.
McAfee, R. Preston; Daniel Quan and Daniel Vincent.
2000. “How to Set Minimum Acceptable Bids, with
an Application to Real Estate Auctions,” unpub.
manuscript.
McAfee, R. Preston and Daniel Vincent. 1993. “The
Declining Price Anomaly,” J. Econ. Theory 60:1,
pp.191–212.
Mei, Jianping and Michael Moses. 2001. “Art as an
Investment and the Origin of the ‘Master-piece
Effect’ Evidence from 1875–2000,” mimeo, Stern
School, NYU.
———. 2002. “Are Investors Credulous? Some
Preliminary Evidence from Art Auctions,” mimeo,
Stern School, NYU.
Milgrom, Paul and Robert Weber. 1982a. “A Theory of
Auctions and Competitive Bidding,” Econometrica
50:5, pp. 1089–122.
———. 1982b. “A Theory of Auctions and
Competitive Bidding, II,” mimeo, Stanford U. and
Northwestern U. Published with new foreword in
The Economic Theory of Auctions. Vol. 2. Paul
Klemperer, ed. 2000. Cheltenham, UK: Edward
Elgar, pp. 179–94.
Natzkoff, Plamen. 2001. “Sequential Auctions: Survey
of the Literature and Empirical Analysis of Auctions
of Raw Wool in Australia,” M.Phil thesis, U. Oxford.
Pesando, James E. 1993. “Art as an Investment. The
Market for Modern Prints,” Amer. Econ. Rev. 83:5,
1075–89.
Pesando, James E. and Pauline Shum. 1996. “Price
Anomalies at Auction: Evidence from the Market for
Modern Prints” in Economics of the Arts: Selected
Essays. in Victor Ginsburgh and Pierre-Michel
Menger, eds. Amsterdam: Elsevier, pp. 113–34.
Pezanis-Christou, Paul. 2001. “Sequential Auctions
with Asymetric Buyers: Evidence from a Fish
Market,” mimeo, econ. dept., Bonn U.
Pommerehne, Werner W. and Lars P. Feld. 1997. “The
Impact of Museum Purchases on the Auction Prices
of Paintings,” J. Cult. Econ. 21:3, pp. 249–71.
Richardson, Andrew. 1992. “An Econometric Analysis
of the Auction Market for Impressionist and Modern
Pictures, 1980–1991,” senior thesis, econ. dept.
Princeton U.
Ross, Myron H. and Scott Zondervan. 1989. “Capital
Gains and the Rate of Return on a Stradivarius,”
Econ. Inquiry 27:3, pp. 529–40.
Stein, John P. 1977. “The Monetary Appreciation of
Paintings,” J. Polit. Econ. 85:5, pp. 1021–35.
Thiel, Stuart and Glenn Petry. 1995. “Bidding
Behaviour in Second-Price Auctions: Rare Stamp
Sales, 1923–1937,” Appl. Econ. 27:1, pp. 11–16.
Thurston, Norman. 1997. “Explaining the Declining
price Anomaly: Empirical Evidence on Buyer
Heterogeneity,” work. paper, Brigham Young U.
Vanderporten, Bruce. 1992a. “Strategic Behaviour in
Pooled Condominium Auctions,” J. Urban Econ.
31:1, pp. 123–37.
———. 1992b. “Timing of Bids at Pooled Real Estate
Auctions,” J. Real Estate Finance Econ. 5:3, pp.
255–67.
Vincent, Daniel. 1995. “Bidding Off the Wall: Why
Reserve Prices May be Kept Secret,” J. Econ. Theory
65:2, pp. 575–84.
Watson, Peter. 1992. From Monet to Manhattan: The
Rise of the Modern Art Market. NY: Random House.
Ashenfelter and Graddy: Auctions and the Price of Art 787