Why study art auctions

1
Why study art auctions?

• Value of most important works of art is
  established by auction
• How the auction system works is a determinant of
  the valuation of art
• Efficiency of the auction system is key
  determinant of cost of creating and distributing
  works of art
• Art auctions provide data that may be used to
  test and refine strategic models of behaviour.

2
Outline

• Mechanics of the auction system
• What we know about how the auction mechanism
  influences price
• What we dont know about how the auction
  mechanism influences price
• How to construct an art price index

3
The Mechanics of Art Auctions

• Major auctioneers are English houses of Sothebys
  and Christies
• English or ascending price auctions
• Sellers set secret reserve prices
• Pre-sale catalogues are published with a low and
  a high price estimate
• Houses earn income from commissions charged to
  buyers and sellers

4
What do we know about how the auction mechanism
influences price?

• Anchoring on Previous Auction Prices
• Effect on Price of Failed Paintings
• The Masterpiece Effect
• The Incidence of Buyers Commissions
• The Declining Price Anomaly

5
Anchoring on Previous Auction Prices

• Suppose two Matisse paintings with identical
  characteristics were both put up for sale.
  However, one of the paintings was previously sold
  in a hot market for a high price and the other
  painting was previously sold in a cold market
  for a low price. Would the painting that
  previously sold in the hot market fetch a
  higher price at auction?

6
Do buyers (or sellers or auctioneers) anchor on
previous prices?

• Necessary to separate anchoring from rational
  learning
• Methodology
• Develop a hedonic prediction of price for both
  the current sale and the previous sale
• Regress sale price on hedonic prediction,
• the difference between the actual price in the
  previous sale and the hedonic prediction, (this
  tests for anchoring)
• and the difference between the actual price in
  the previous sale and the hedonic prediction in
  the previous sale (this expression controls for
  unobservable charcteristics)

7
Estimating equation
8
(No Transcript)
9
(No Transcript)
10
Results

• A 10 percent positive difference between the
  previous sale price and the hedonic prediction
  would lead the final price to be adjusted upward
  toward the previous price (the anchor) by between
  6.2 percent and 8.5 percent of this difference
  depending on the specification, with symmetric
  results for negative differences.
• For Contemporary art, the adjustment is between
  0 and 50, depending upon the specification.
• We find similar results using the low estimate.

11
Interpretation

• Strong support for anchoring effects.
• We interpret these effects as anchoring on the
  part of the buyers, with the sellers and
  auctioneers either anticipating anchoring on the
  part of the buyers or exhibiting similar
  anchoring effects themselves.
• We found no evidence of asymmetric effects
  between gains and losses.

12
Are Bought-in Paintings Burned?

• Often claimed when an advertised item goes unsold
  at auction, its future value will be affected.
• Cristallina case (1981) 7 out of 8 paintings
  failed
• Did a lack of care result in a loss of value?

13
How to Test

• Use a repeat-sale data set in which all
  paintings have been sold twice
• Some of the items appear and fail between sales,
  others do not
• Then test for return differences between
  paintings that failed and paintings that did not

14

• Regression
• Return is conditioned on whether or not a
  painting appeared and failed
• Items that failed to sell return 30 less than
  other paintings

15
(No Transcript)
16
Why are paintings burned

• Trend effects failure is correlated with
  downward trend in tastes
• Reserve price effects A seller lowers the
  reserve price after a failure because of an
  urgent need to sell.
• Common value effect Failure causes a drop in
  price because bidders take into account other
  bidders valuations of the painting

17
(No Transcript)
18

• Important Old Masters I
• New York, Jan 26, 2005, Lot 5
• Luigi Miradori, il Genovesino 1600-1656
• Portrait of an Olivetan Priest
• 100,000-150,000
• 142.5 x 97 cm.

19
(No Transcript)
20

• Important Old Masters I
• New York, Jan 26, 2005
• Lot. 38 Antonio Belluci 1654-1726
• The Triumph of Galatea
• 50,000-70,000 50x51 cm
• Lot. 31 Sebastiano Ceccarini 1703-1783
• Portrait of a Child, said to be Marie-Louise of
  Parma
• 60,000-80,000 104.8 X 80.7 cm

21
The Masterpiece Effect

• Art dealer Edward Merrin
• Its 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.
• Most authors have tested for the Masterpiece
  Effect by constructing a portfolio of the top 10
  or 20 of prints by price. If estimated price
  indices in this portfolio outperform another,
  then there is a Masterpiece Effect.
• Generally, various authors have not found support
  for a Masterpiece Effect, and in general have
  found that Masterpieces underperform other
  portfolios

22

• A negative masterpiece effect may be due to
• overbidding and then mean reversion
• Reasonable, especially given definition of a
  Masterpiece
• survivorship bias
• More expensive paintings remain in the sample
  throughout, whereas less expensive paintings drop
  out
• Are economists really measuring masterpieces
  correctly?

23
(No Transcript)
24
Who Pays the Buyers Commission? Background

• Christies and Sothebys were found guilty in the
  US of price fixing
• The related class action by 130,000 of the
  houses customers resulted in a settlement of
  512 million
• Christies and Sothebys charge commissions both
  to sellers (sellers commissions) and to buyers
  (buyers premiums)
• They should compete on these dimensions
• Impressive evidence of secret agreements to set
  the level of commissions charged to sellers

25
Settlement Details

• Sellers are assumed to have been overcharged 1
  percent and buyers 5 percent of the sales price
  on each purchase of 50,000 or less, with a
  maximum of 2,500 for each purchase above 50,000
• Most probably, sellers will receive a third and
  buyers two-thirds of the 512 million
• Applies only to buyers and sellers in successful
  bids in New York auctions

26
Does this make sense?
27
Some Auction Theory

• The maximum amount a bidder is willing to bid is
  his reservation price
• In an oral ascending auction, the price paid is
  the reservation price of the second highest
  bidder
• If the commission rate increases, the actual bid
  (before commission) will decrease by the same
  factor in order to keep the effective price paid
  by the buyer equal to his reservation price
• No logical reason to compensate the buyers

28
Who should be compensated?

• Sellers lose since they receive a lower price for
  their object when commissions rise
• This lower price could discourage sellers from
  bringing their paintings to market
• Welfare is decreased
• Not taken into account in settlement as only
  successful buyers are compensated

29
The Declining Price Anomaly

• First publicized by Ashenfelter (1989) for wine
• Prices are twice as likely to decrease as to
  increase for identical bottles of wine sold in
  same lot sizes
• First noted by Penny Burns (1985) for wool
  auctions
• A large theoretical literature as to why it might
  occur
• A large empirical literature showing declining
  prices in many types of auctions (including art
  auctions), and rising prices in some auctions

30
The Declining Price Anomaly

• Wine auctions
• Prices are twice as likely to decrease as to
  increase for identical bottles of wine sold in
  same lot sizes
• Condominium auctions
• Controlling for later sale price, one study found
  approximately a 5.2 decline throughout the
  auction

31
Why is this an Anomaly?

• First object has n bidders, second object has n-1
  bidders
• With fewer bidders, prices should decline,
  however.
• Bidders shave bid for first object in
  anticipation of second object
• If items were identical, price of first object
  should equal price of second object

32
The Declining Price in Art Auctions

• Contemporary Art
• Pre-sale estimate decreases about .5 per item
  sold or 30 throughout the auction
• Controlling for pre-sale estimate, bid decreases
  about .09 per item sold or about 6.3 throughout
  the auction

33

• Impressionist and Modern Art
• Pre-sale estimate decreases about 1.2 per unit
  sold or 92 throughout the auction
• Controlling for pre-sale estimate, bid still
  decreases about .05 per unit sold or 3.5
  throughout the auction

34

• London, Thursday, 5 October 1995
• Chateau Petrus -- Vintage 1961
• Lot 105 impeccable appearance
• Lot 106 damaged label. Excellent level.
• Lot 107 good label. Level base of neck.
• Lot 108 slightly damaged label. Level base
• of neck
• Lots 109 and 110 good labels. Levels top
• shoulder.
• Lot 111 slightly damaged label. Level top
• shoulder.

35
Our Explanation

• Even though bidders should shave bid for first
  object in anticipation of the second object, with
  nonidentical items, if first item is preferred,
  do not shave bid enough, i.e. even controlling
  for pre-sale estimate, price will be higher for
  first item

36

• Black and DeMeza (1992) claim they exist
  primarily because the winner of the first auction
  in a sequence has the option to buy the remaining
  objects at the winning price
• McAfee and Vincent (1993) showed that risk
  aversion could create declining prices
• Von der Fehr (1994) whows that participation
  costs could create declining prices through
  strategic bidding
• Ginsburgh(1998) shows that the presence of
  absentee bidders can generate declining prices
• Declining prices found in auctions in a
  tremendous number of different types of items
• A lot (too much?) has been written on declining
  prices

37
What we dont know about how the auction
mechanism influences price

• Do Auction Houses Provide Truthful Information
  about the Items Being Sold?
• Theory suggests they should
• A (fairly) large number of papers with
  contradictory results

38
(No Transcript)
39
Is more research necessary?

• Does it matter?
• Mei and Moses (2005) Auction house estimates
  are effected by agency problems AND some
  investors are credulous
• Beggs and Graddy (2009) It is possible (though
  no supporting evidence) that buyers and sellers
  anchor on pre-sale estimates
• (More probable that both buyers and auctioneers
  anchor on previous prices).

40
How to Construct a Hedonic Art Price Index

• Items are typically unique
• Average prices may be exacerbated during booms as
  better paintings may come up for sale
• Auction indices are based on the model

41
Two Rats, Vincent van Gogh
42
Portrait of Dr. Gachet, Vincent van Gogh 82.5
million
43

• Hedonic Models
• Advantage All the data may be used in the
  estimation, including data on objects that are
  only offered once in the sample period
• Disadvantage The strong assumption that a small
  set of x variables captures much of the
  variability in the fixed components of price
• Systematic movements in the unobserved
  characteristics of the objects being offered for
  sale may bias the results.
• Nonetheless, these models have been used
  extensively in estimating the returns to art (see
  table)

44
Repeat Sales Models

• Introduced by Baily, Muth and Nourse (1963) for
  real estate
• Improved upon by Case and Shiller (1987) for real
  estate
• Used extensively for art and cultural items (see
  table)

45
How to Construct a Repeat Sales Price Index

• Based on premise that
• ri,tµt ei,t
• ri,t is the continuously compounded return for
  a particular art asset i in period t, usually
  constructed as ln psale- ln ppurchase
• µt is the average return in period t of paintings
  in the portfolio
• usually estimated by constructing either 1, -1
  dummy variables (if an index is being estimated),
  or dummy variables equal to 1 during the holding
  period and 0 otherwise
• ei,t is assumed to be uncorrelated over time
  and across paintings
• Note that this model can be derived from a
  general hedonic model above by differencing
  identical paintings

46

• Thus, summing over time, the logged ratio of the
  sale price to the purchase price can be expressed
  as
• Baily, Muth and Nourse estimated this by OLS.
• Case and Shiller noted the presence of
  heteroskedasticity relating the the holding
  period (evidenced by the sum of the error terms
  above)

47
Case and Shiller 3-stage least squares

• First stage OLS
• Second stage squared residuals from OLS are
  regressed on duration (time held) and a constant
• The slope coefficient indicates the degree to
  which the error increases with the number of time
  periods the painting is held
• The constant is interpreted as an estimate of the
  variance in the painting-specific error term
• Third stage a generalized least squares
  (weighted) regressions is run that repeats the
  stage-one regression after dividing each
  observation by the square root of the fitted
  value in the second stage.
• The Case and Shiller method is correct as long as
  the error terms are uncorrelated over time and
  across paintings

48
Repeat Sales Models

• Disadvantage Based on a small number of
  paintings
• Paintings that are sold twice may be different
  i.e. survivorship bias
• Problems with both methods General only use
  sold items

49
(No Transcript)
50

• Correlation of indices .9559 std. dev. of
  hedonic index1.025, std. dev. of repeat sales
  1.166
• Real return on hedonic index 4
• Real return on repeat sales index is 1.166
• Two different explanations
• 1. Major impressionist sales are held in October.
  Hedonic index underestimated returns for this
  short period of time, because it was unable to
  correct for quality differences early in the
  year.
• 2. Because the repeat-sales index is based on
  such a small number of paintings during that
  period, these paintings were unrepresentative
• Hedonic model has 8792 observations whereas the
  repeat sales model has only 474 observations
• These criticisms are not applicable to prints

51
(No Transcript)
52
(No Transcript)
53
Conclusions

• Economists are beginning to understand some of
  the influences that the auction mechanism has on
  art prices
• An introduction of psychological behavior into
  the standard models might be helpful
• Data from art auctions can be used to test
  economic theory that is widely applicable to a
  variety of situations