Buying paintings by numbers

Subscribe to Artprice.com or Artnet.com and they will show you every work any given artist has sold at auction for the past 10 years or more. You can find out where the work was sold, the estimate, and what it sold for (including whether it failed to sell).

You can also have those numbers crunched into charts that show the average prices paid for the artist’s work in any year, the volume of works sold and all kinds of other metrics. It is almost like playing with bond prices on a Bloomberg terminal.

The phenomenon is part of art’s emergence as an alternative asset class, right up there with stocks, bonds and real estate. That has brought with it all the trappings of professional investment management – including databases such as artprice.com and, almost inevitably, indices designed to track the art market.

There are now about half a dozen businesses offering art market indices that track the “performance” of artists and genres of art over time. You can get a Picasso index, an Old Masters index, or even – courtesy of Art Market Research – indices on collectibles such as teddy bears and fountain pens.

But can these indices really work as well as those for altogether more liquid investments such as stocks and bonds?

Well-established dealers and art lovers view the trend, which appears to be gaining momentum, with distinct unease. But the transparency that computers and databases have brought to the market is hard to deny. A collector now has ready access to a vast quantity of information about the prices paid for a particular artist’s work, where prev­iously he or she was reliant on a dealer’s experience and knowledge.

Many art buyers are from the corporate or financial world and have taken readily to the indices and data services. They are accustomed to crunching numbers and often feel more comfortable amassing such quantit­ative data when considering a purchase.

The indices are also used by a handful of art funds that have been launched in the hope of attracting money from investors eager to reap returns from the new asset class. Shares in Artnet.com, probably the biggest purveyor of online art data, have risen by 600 per cent in the past two years as the art market has boomed. It has more than 5,000 subscribers and its website has more than 1m unique users a month.

But there are pitfalls in using the indices as a proxy for the actual price a work will fetch. One big problem is that most indices simply show the average price paid for an artist’s work over any given period. But works of art, unlike stocks or bonds, are not interchangeable. One Microsoft share is the same as another, so a graph of the Microsoft share price shows you what price you are likely to get for your shares.

But one Picasso is not the same as another. An index of Picasso’s works sold in the past year might include a particularly desirable work that went for $80m, which would skew the results upwards.

Or a collector might have sold an unusually large number of minor works that did not fetch much – causing the “index” for that year to fall, giving a misleading impression that Picassos in general are out of favour. Such phenomena also exist to some degree in the stock and bond markets but the much greater volume in those markets means the effect is more muted.

Most art index providers get around these problems by “trimming” or smoothing the index – removing, say, the top 5 per cent or bottom 5 per cent of works, as measured by price. Sometimes they might subjectively decide a work should not be included. In other words, most of the indices are not really indices but graphs of average prices. That in turn ignores the central fact that artworks are not interchangeable.

“Art data is not suitable for promoting art as an investment,” says Kevin Radell, in charge of financial products at Artnet.com. “Most indices aren’t meaningful as price/performance guides. They can tell you something but they cannot be considered a reliable guide in the same way as a stock or bond index.

“To pretend average price indices are the same as price/performance indices is alchemy. They are measuring different paintings, so they are meaningless from a financial point of view.”

Artnet.com prefers to refer to its price charts as “normalised market data”, although Radell is attempting to compile true indices. For this, some way of reliably quantifying the judgment used by art appraisers – measuring the condition and the rarity of the artwork, for instance – would have to be found and incorporated into the data.

Radell, a former investment banker, says the art market is full of price inconsistencies, which further diminish the value of indices as price indicators. Some works and genres will sell very well in certain countries but not in others. Poss­ibly the best known index
in the art world is the Mei Moses index, created by Mike Moses and Jianping Mei, New York University academics. It uses repeat sales only – that is, it includes only those works for which the previous sale price is known. Their index comes closest to a real index in that it reflects actual increases in value, not general averages.

But there is naturally a much smaller number of transactions underlying their index. Also, it uses data only from Christie’s and Sotheby’s New York auctions.

Moses has just completed a study comparing his repeat sale index with an average sale index – with interesting results. Taking data based on auction sales of Impressionists from 1995 to 2004, he created an average price index and a repeat sales index. The two moved in roughly the same direction, ending the decade higher than they started. But the average sales index was far more volatile than the repeat sales index. In one year, 1997, it rose by almost 400 per cent before plummeting in 1998.

In the same period, the Mei Moses repeat sale index barely moved, showing a startling difference between the two indices, even though they were based on the same data. “The average index implies you are getting annual returns that are not actually achievable in the marketplace,” says Moses.

Another problem for all the indices is that the databases use only auction sale prices. None of the “indices” includes dealer prices (although Artnet.com is adding dealer prices to its database). Since about two-thirds of all transactions are made through dealers, this makes for incomplete data. It would exclude, for example, Ronald Lauder’s recent purchase, for $135m, of a work by Gustav Klimt (“Adele Bloch-Bauer I”), probably the most expensive work yet sold.

The Klimt “index” would look very different if that transaction were included. Also, some artists, for whatever reason, do not sell much at auction but have a strong following through dealers. Their price “index” could be quite misleading.

None of the indices takes into account buy-ins or works that fail to sell because they do not meet their reserve price. This can comprise a third of the works put up for auctions. Excluding buy-ins this means most of the indices are almost certainly skewed too high.

Even using auction prices, there are discrepancies among the databases. A recent study provided by Artnet.com showed 284 Artnet auction results for the American artist Edward Hopper. Artprice.com, its main competitor, had 173 results for Hopper; Artfact.com had 119 and Askart.com, 105.

For Marc Chagall, Artnet had 19,215 sale results; Artprice.com had 10,449; Artfact.com 3103, and Ask art.com had 27. Results for several dozen more artists showed similarly wide ranges, even though the databases mostly use the same auction house results.

All this is not to say that the art databases and indices are of no use. They provide a depth of information and transparency that is welcome to curators and collectors. In aggregate, they also provide a general pointer on where the market is heading. But if the average price of Old Masters sold has risen by 50 per cent in a year, that does not mean that if you bought an Old Master last year, you can now sell it for 50 per cent more.

Caveat emptor.