In 2016, an index tracking sales of fine art objects at major auction houses was sold to Sotheby’s (one of the houses being tracked) for an undisclosed sum. This index tracks what are called “same sale pairs” meaning a painting that is sold twice or more times, allowing its price performance to be tracked. At the time of the acquisition the index was comprised of 45,000 such pairs; up from an initial 4,896 analyzed in the original 2002 paper we are analyzing today from economics professors Jianping Mei and Michael Moses.
This section aims to point out the issues around the dataset this study is built upon. The core criticism that you will see of this paper is one of a lack of substantive data. You see, the art market is a very opaque enterprise; though the two major houses being examined here, Christie’s and Sotheby’s, dominate the auction market, they only account for a percentage of all fine art sales. Galleries, private dealers, direct-from-artist prices all vary wildly due to the environment wherein the purchase takes place. The authors make sure to note the narrow purview of their dataset.
This is not to say that the data is worthless, studying some percentage of the market can still give us a better understanding of the market as a whole. Though, there are a few other issues in the data that the authors spend this section pointing out.
Most importantly, the authors call attention to the potential selection bias in the study. In the context of this paper, selection bias comes in the form of the data set being based on sales from major U.S. auction houses, which tend to truncate both sides of the return distribution. It is also pointed out that missing is data on property that went up for sale, but did not meet the reserve price(1). Another significant factor mentioned by the authors is that of donations—uber high quality works are often given to museums from collectors (for among other reasons, a tax write-off), further skewing the data.
Mei and Moses also mention a potential backward-filled bias(2) that may appear in their findings. The fact that the data from pre-1950 is populated by well known auction house sales may skew returns upward due to these sales being made up of already-established artists from the time period.
Knowing this, we must understand that all manner of biases affect financial markets. So while we should not take the conclusions of this paper wholesale, there is absolutely knowledge to be gained from its findings.
Section two discusses the repeat-sales regression (RSR) method, which is used to estimate the fluctuations in the value of an average asset (painting in our case) over a particular time period. The benefit of using RSR in this case is that it controls for differing quality of assets; a helpful mechanism considering the differences in quality of paintings in the dataset. However, as mentioned earlier, the data used here is beset with data biases. So we must take caution when extrapolating from these findings.
The rest of this section outlines the statistical models used to build the index. We won’t dig into the mathematical nitty-gritty, but a couple of things are important to note.
The authors use a generalized least-squares regression to estimate the average return of a portfolio of paintings. By analyzing price changes in an individual work we are able to control for differences in quality among paintings; meaning high performing repeat sales of a high quality work will not affect the perceived quality of another work—in the eyes of the model that is. Said another way, each artwork is valued based on its own price history, allowing researchers to assume that price movements result from changes in market factors as opposed to changes in the quality of a work itself.
The price history of the paintings being tracked goes from 1875 to 2000. Over different periods within this time frame paintings performed admirably against the other financial instruments the authors decided to track.
From 1950-1999, the art index had an annual compounded rate of return of 8.2%, which is about on par with the 8.9% of the S&P and the 9.1% of the Dow during the same run. It outperformed both corporate and government bonds, which earned 2.2% and 1.9% respectively.
From 1900-1999, the Mei-Moses index underperformed stocks much more noticeably. While art earned a 5.2% return, the S&P and Dow earned 6.7% and 7.4% in the same period. The art index again outperformed both corporate and government bonds—which only received 2.9% and 1.4% over the century time frame.
From 1875-1999, the art index returned 5.2%. Losing out again only to the two stock indices: S&P earned 6.6% while the Dow earned 7.4%.
One important point to take note of is that over all these time periods art had by far the highest standard deviations(3) in each. Ranging from a low of 0.213 in the 1950-1999 period to a high of 0.428 in the 1875-1999 period. Each of the standard deviations for the paintings was higher than those for any of the other investment products on view.
It would seem that paintings as an investment over the period would favor a similar clientele that they do today. A well-moneyed group that is not put off by an asset class with high volatility; even if the payoff is only mediocre to good returns at best.
Common, albeit self-serving, advice from art dealers to invest in the most expensive works one can afford is rigorously analyzed in section four of the paper. It is a common sentiment among art market professionals that pieces with the most robust returns are more likely than not “masterpieces”. There had been previous analysis of this hypothesis by other researchers, but Mei and Moses were able to dig into this thesis specifically focused on expensive paintings due to their unique dataset.
What they found is surprising, considering it goes against what still is a pervasive opinion in the fine art market. They conclude that masterpieces did in fact underperform the broader art market.
In the table below, 'y' shows the difference in returns between masterpieces and the overall art market, considering their purchase prices. A negative 'y' means masterpieces perform worse than the general market. The t-statistic measures the reliability of the 'y' value. A higher absolute t-statistic means the observed difference in returns is less likely to be random, and more likely to be due to the masterpiece status of the work.
They built an equation that models the elasticity of art returns with respect to the log price of the property. Said more simply, the equation is a measure of the expected change in annual returns per some change in purchase price. Similar to the way one would model expected returns of buying shares in a high or low market capitalization of a company.
The authors even mention a similar “masterpiece underperformance” phenomenon within the stock market. The so-called “small firm effect” is an explanation for why smaller companies may achieve excess returns not justified by their risk profile.
Another potential explanation proffered by the authors is one of the purchase of art for pleasure or personal consumption. Since extremely high quality works have a high chance of ending up in penthouse apartments and gaudy mansions, it would stand to reason that utility the piece is generating is one of personal satisfaction instead of investment returns. Also, that when a piece experiences extraordinarily high priced transactions its return profile may be dampened by the future return reverting to the mean instead of continuing on an accelerated trajectory.
One issue we take with this section is how the authors define a masterpiece. The paper tracks masterpieces in so far as they track the returns of a work per its purchase price, but according to many this would not guarantee the work is truly a masterpiece.
For example, if you spent $5 million on a George Condo canvas you likely received a masterpiece. The artist’s most expensive public sale was for around $6 million, meaning that it would be safe to assume most works of his around that price point could be assumed a masterpiece. At least, in the eyes of high net worth collectors that is—I digress, though.
But if you were to buy a Pollock or Monet canvas today for $5 million, then the definition does not necessarily hold true. A Painting being expensive alone does not a masterpiece make. Assessing the elasticity of returns based on purchase price is a worthwhile analysis. Nevertheless, evaluating the performance profiles of masterpieces as a whole necessitates examining the œuvre(5) of each artist, which may problematically be subject to the analyst's personal taste.
The “Law of One Price” is an economic maxim that states "In the absence of trade frictions, and under conditions of free competition and price flexibility, identical goods sold in different locations must sell for the same price when prices are expressed in a common currency". An obvious challenge to testing this principle within the market for paintings is that an identical object is only sold a handful of times, usually over very many years. It cannot, for example, be sold at Christie’s and Sotheby’s separately on the same day.
Economist James Pesando analyzed this phenomenon in 1993, but his dataset was comprised of prints—a medium that does allow for two identical(4) work’s prices to be checked against one another. He found that the law is broken due to prints regularly selling for higher prices at Sotheby’s as opposed to Christie’s.
Mei and Moses found mixed results in this situation. 'Place of transaction' for American paintings seemed to have no bearing at all on performance or prices. Impressionist works on the other hand, experienced statistically significant higher returns when they were bought at other auction houses and ultimately sold at Sotheby’s; this would indicate that Sotheby’s did, on average, fetch higher prices during the period. Furthermore, when a work was bought at Sotheby’s and then sold at Christie’s it tended to receive lower returns compared to those paintings that were both bought and sold at Sotheby’s.
In the table below, 'p' represents the difference in returns between each buy-sell combination of houses. A positive 'p' means that the specific buy-sell combination outperforms the benchmark (both buying and selling at Christie's), while a negative 'p' indicates underperformance against the benchmark. The t-statistic measures the reliability of the 'p' value. A higher absolute t-statistic suggests the observed difference in returns is less likely to be random, and more likely to be dependent upon the "Buy—Sale" pair of the work.
This is all to say that the best way, according to the paper, to achieve quality returns with respect to the law of one price is to buy at smaller auction houses and then sell at one of the majors. The blue-chip reputation of the big houses lends credence to this theory.
The authors close the paper by restating the fundamental findings of the paper. Namely that art has outperformed some fixed income securities (and that gives a reason why portfolio diversification may be aided by the addition of artwork), so-called “masterpieces” tend to underperform the art market, and finally that the “Law of One Price” may or may not have a statistically significant effect on the New York auction market.
They go on to mention again that sample selection bias may be injecting significant bias into the return estimates.
The paper closes with a number of interesting questions that go unanswered: is there a systemic bias in bidding prices so that winning bids tend to exceed true value? Would a time-series analysis of this data imply that the entire market regularly experiences a mean reversion process? They kindly leave these questions for future research and we offer them to you, reader.
Photo-Credit: Tingey Injury Law Firm
Disclaimer: You understand that by reading Altan Insights, you are not receiving financial advice. No content published here constitutes a recommendation that any particular security, transaction, or investment strategy is suitable for any specific person. You further understand that the author(s) are not advising you personally concerning the nature, potential, value or suitability of any particular security, transaction, or investment strategy. You alone are solely responsible for determining whether an investment, security or strategy, or any other product or service, is appropriate or suitable for you based on your investment objectives and personal financial situation. Please speak with a financial advisor to understand if the risks inherent in trading are appropriate for you. Trade at your own risk.