What Sotheby’s Mei Moses Doesn’t Tell You
Analyzing art as an investment has two major obstacles. The first is the relative lack of data, as art sales happen much less frequently than financial trades. The other is that every piece of art is different, so it becomes difficult to compare how one piece sells with respect to another. One way to develop a view of the market despite these challenges, is to develop an index, a set of values over time that show which way the market is trending. An index can verify trends and track volatility of the market. Two markets with vastly different capitalization are often compared using their indices. Finally, fund managers can use an index as a benchmark for their fund performance, i.e. if they are vastly underperforming the market, they likely want to change strategies. In this post, we consider how indices in the art market are constructed.
To build an index, two major classes of mathematical models have been developed for predicting prices: 1) the repeated sales method, and 2) hedonic regressions.[1] Arthena uses a hedonic regression to predict artwork values for a number of reasons beyond just building indices, but it is worth comparing this choice to the art world’s premier index, the Mei Moses Art Indices, which uses the repeat sales method.
Jianping Mei and David Moses, two NYU Stern professors, developed this index, and released it in 2001. At that point, their data set was 4,500 hand picked pairs of sales representing one work’s acquisition and resale at a later date. Initially the index had an annual basis, but with more data, they are able to calculate a monthly index. When Sotheby’s acquired the index in 2016, that number had grown to 45,000, and the makers of the tool claim that the number of repeat sales added each year is around 4,000.[2]
The major disadvantage of a repeat sales method, is that only data representing repeat sales can be included. In real estate, 90% of sales represent properties being resold, but in artwork, less than 15% of sales are repeat sales.[2] Therefore, the repeat sales method only uses a portion of available data. This has the potential to introduce a large bias into the index, as these 15% of sales may not be representative of the market, considering that works that are particularly good or bad are more likely to be resold. [3]
Mei and Moses state that the major advantage of the repeated sales method is that it does not suffer from the ‘arbitrary’ specification of a hedonic regression. Our belief is that the hedonic regression can be carefully designed to maximize its predictive power, and that that outweighs any sense it might be less objective. Keep in mind that both models produce errors in their estimates, therefore it is up to the modeler to reduce this error as much as possible and choose which is more useful to them.
For the curious, a repeat sales index starts with the assumption that while a work is held, its value compounds continually at a rate R. This rate is calculated easily as the natural logarithm of the ratio of the sale price to the acquisition price. R can be imagined to be the sum of yearly rates, r, for the work for each of the years it was held. In turn, every year has an average return rate, m, across all works that were held that year. The goal is to find the set of m’s that minimize the total difference between all the m and the active r’s for each year. With the m in hand, an index can be calculated by setting the first year to some base value, usually 100, and calculating new values according to the growth rate specified by each year’s m. [4]
While an index calculated this way can tell you which way the market is trending and how volatile it seems, it leaves much to be desired as an investment tool. The Mei Moses Art Indices can show you that you may want to invest in art, or a certain sector of art, but moving on to choosing a particular piece of work is a shot in the dark. Arthena’s methodology, which uses hedonic regression, answers these questions, which we will explore in future posts.
[1] Bocart Fabian Y. R. P., and Hafner Christian M. (2015), Volatility of Price Indices for Heterogeneous Goods with Applications to the Fine Art Market, J. Appl. Econ., 30, pages 291–312. doi: 10.1002/jae.2355
[2]:Artnet — See What Experts Have to Say About Sotheby’s Acquisition of the Mei Moses Art Indices
[3] Chanel, Olivier, Louis-André Gérard-Varet, and Victor Ginsburgh. “The Relevance of Hedonic Price Indices.” Journal of Cultural Economics 20, no. 1 (03, 1996): 1–24. doi: http://dx.doi.org/10.1007/s10824-005-1024-3.
[4] Mei, Jianping, and Michael Moses. “Art as an Investment and the Underperformance of Masterpieces.” The American Economic Review 92, no. 5 (2002): 1656–668. http://www.jstor.org/stable/3083271.
