|
Article Excerpt
I. INTRODUCTION
Financial assets such as stocks and bonds are traded almost continuously, while the sales of art objects are episodic and often infrequent. The most significant challenge to those seeking to estimate the risk and return of investing in art is the need to track the market price of identical works of art over time. As first noted by Pesando (1993), the frequency of repeat sales makes the market for prints an excellent laboratory in which to study the investment performance of art. Prints are multiples, typically published in editions of 50, 100, or more, and the result is to increase dramatically the number of repeat sales available to estimate an index of art prices. The sample period in Pesando reflects the availability, commencing in the year 1977, of comprehensive data from Gordon's Print Price Annuals on prices realized at all of the world's major auction houses. The end of the sample, the year 1992, reflects the date of the study. Pesando concludes that, on a risk-return basis, an investment in modern prints is inferior to an investment in traditional financial assets. Contrary to the conventional wisdom of the art trade, higher priced prints--masterpieces--tend to underperform the market as a whole.
With the availability of 12 yr of additional data, it is an opportune time to reexamine these earlier findings and to situate these updated findings into the rapidly expanding literature on the performance of art as an investment. In addition, with increased data, we are able to examine the performance at auction of individual artists. This, in turn, permits us to explore the possible impact of changing tastes on investment performance.
The paper proceeds as follows. We estimate the risk and return of investing in modern prints for the extended sample period 1977-2004 and compare this performance with investment in traditional financial assets as well as gold. In so doing, we uncover a puzzle: the fact that the returns on prints sold at the major venues--Sotheby's and Christie's in New York and London--are much higher than those on prints sold at all auctions worldwide. We then reexamine the "masterpiece effect," the claim that higher priced works of art will outperform the market as a whole, including lower priced works of art. As emphasized by Baumol (1986), the price of a work of art is not anchored by a future stream of payments, unlike a traditional financial asset, and may fluctuate in response to essentially random change in tastes. With this in mind, we examine the relative performance of investment in the modern artists whose prints are most frequently sold at auction: Picasso, Chagall, and Miro.
II. THE DATA SET
Modern prints are original graphic works (i.e., etchings, drypoints, lithographs, etc., in which the artist conceived and executed the original image as a print) by artists such as Picasso, Chagall, Miro, and Matisse. Prints of these artists have a long history of sale at auction, and price differences due to "nonobserved" variations in condition and quality are much less important than in the market for old master prints (Rembrandt, Durer, etc.).
Because prints are multiples, different impressions of the same print frequently appear at auction. As a result, one can use repeat sales of the same print to overcome the fundamental problem faced by those attempting to estimate an index of art prices. This problem is the heterogeneity of art objects, which makes it difficult to track the price of an "identical" art object over time.
Gordon's Print Price Annual was published for the first time in 1978 and has been published in every year since. Each Annual contains a complete record of the prints sold at the world's major auction houses (Sotheby's, Christie's, etc.) during the previous year. Each print is identified by the artist and catalogue raisonne number. The Annual also indicates whether the print has been signed by the artist, is unsigned, or has a stamp signature. The Annuals, compiled from the catalogues prepared in advance of the sales and the price lists released
thereafter, currently contain more than 50,000 entries. Prices are inclusive of the buyer's commission. If the auction did not take place in the United States, prices are converted to U.S. dollars at the then exchange rate.
With the passage of time, the number of artists and the number of auction houses included in the Annuals have increased. We continue to focus on the 28 artists used in the original study by Pesando. However, to be comprehensive, we include the additional auction houses that now sell modern prints. To check the robustness of our findings, we also estimate a price index for prints sold at the four most significant auction venues (Sotheby's and Christie's in New York and London).
III. THE INDEX OF PRINT PRICES
To estimate the index of art prices, we use the repeat sale regression method first advanced by Bailey et al. (1963). A repeat sale occurs whenever the "identical" print (i.e., artist, catalogue raisonne number, signed, or not) is sold on two different occasions. For each pair of sales, the log-price relative is calculated: the log of the price on the later sales date less the log of the price on the earlier date. The log-price relatives are then regressed on a set of dummy variables, one for each observation of the log-price index. For each observation of the dependent variable, the dummy is set equal to + 1 at the time of the second sale, - 1 at the time of the initial sale, and at all other times. If the initial sale is in the first time period, there is no dummy variable corresponding to the initial sale. Since auction sales for modern prints are concentrated during May-June and November-December, we work with semiannual observations.
The repeat sales method yields an estimate, as well as the standard error, of each value of the log-price index. The result is an index of nominal prices, which is readily converted to real prices using data on the U.S. consumer price index.
The repeat sales method is one of two methods used by researchers to solve the essential problem in estimating indexes of art prices: the need to track, over time, the price of an identical art object. The alternative approach is hedonic price models, which seek to hold constant the key characteristics of each work of art by including a small number of hedonic characteristics, such as the name of the artist, the medium, the size of the artwork, and so on. The advantage of the hedonic models is that all observations on realized prices can be included in the estimation of the price index. The disadvantage is that the small number of hedonic characteristics is incapable of capturing the true quality of an artwork, which may lead to biased estimates of the price index. Most obviously, and as noted by Ashenfelter and Graddy (2006), the aesthetic merit of an...
|
