Művészet | Festészet » Mei-Moses - Economic Activity and Painting Performance

Source: http://www.doksinet Economic Activity and Painting Performance Jianping Mei Department of Finance, Stern School of Business, New York University, 44 West 4th Street, New York, NY 10012-1126 jmei@stern.nyuedu Michael Moses 1 Stern School of Business, New York University, 44 West 4th Street, New York, NY 10012-1118 mmoses@stern.nyuedu August 2002 Keywords: Fine Art Index, Art Performance 1 We have benefited from helpful discussions with Will Goetzmann, Robert Solow and Larry White on art as an investment as well as Rohit Deo and Andrew King on the statistical analysis. We would also like to thank Mathew Gee of the Stern Computer Department for his tireless efforts in rationalizing our database. We also wish to thank John Campbell for his latent variable model algorithm. All errors are ours 1 Source: http://www.doksinet 1. Introduction Two major obstacles encountered in undertaking a financial analysis of the art market are heterogeneity of artworks and infrequency of

trading. The present paper overcomes these problems by using a repeated-sales data set based on art auction price records available at the New York Public Library and the Watson Library at the Metropolitan Museum of Art. The data set was first reported in Mei, Moses (2001) and is updated through 2001 for the present study. For the artworks included in the present study, we have over 5,400 price pairs covering the period 1875-2001. As a result, we analyze a significantly larger number of repeated sales than earlier studies by William J. Baumol (1986) and William N. Goetzmann (1993) 2 With a larger data set, we are able to construct separate annual art indexes for American, Old Master and 19th century, and Impressionist and Modern paintings for the time period 1950-2001. These annual indices are then used to address the question of whether the risk-return characteristics of paintings in each of these collecting categories compare favorably to the risk–return Jianping Mei and Michael

Moses. 2 Goetzmann (1993) uses price data recorded by Gerald Reitlinger (1961) and Enrique Mayer (1971-1987) to construct a decade art index based on paintings which sold two or more times, during the period 1715-1986. The Goetzmann data set contains 3,329 price pairs. Baumol (1986) uses a subset of the data recorded by Reitlinger (1961) to study the returns on paintings during the period 1652-1961. The Baumol data set contains 640 price pairs. Buelens and Ginsburgh (1993) re-examine Baumol’s work with different sample 2 Source: http://www.doksinet characteristics of traditional financial assets, such as the S&P 500. The indices are also used to determine whether changes in geopolitical and economic activity cause changes in the these indices The remainder of the paper is organized as follows. Section 2 briefly describes the art auction data set and the repeated sales regression procedure used to estimate the indexes for painting prices 3. Section 3 provides risk and return

characteristics of the estimated art price indexes. Section 4 presents evidence on the effects of wars and recessions on the art indexes. Section 5 tests the hypotheses that oil prices, US inflation or global stock markets influence art prices. Section 6 presents the conclusions of the paper. 2. Painting Data and Computational Methodology Because individual works of art have not yet been securitized and there are no publicly traded art funds, changes in the value of works of art cannot be determined from financial sources. The prices of artworks purchased from a gallery or directly from the artist tend not to be reliable or easily obtainable. Repeat sale auction prices, however, are reliable and publicly available (in catalogues published by the auction houses) and can be used to construct a data base for determining the change in value of art objects over various holding periods and collecting categories. We created such a database for the U.S market, principally from sales in

auction houses in New York City. We searched the catalogues for all American, 19th Century and Old Master, and Impressionist and Modern paintings sold at the main sales rooms of periods. Pesando (1993) uses a database for modern prints which has 27,961 repeat sales, but the data covers only a short time span from 1977 to 1992. 3 For a more complete discussion of these factors please reference Mei, Moses 2001. 3 Source: http://www.doksinet Sothebys and Christie’s (and their predecessor firms) from 1950 to 2001. 4 If a painting had listed in its provenance a prior public sale, at any auction house anywhere, we reviewed that auction catalogue and recorded the sale price. Some paintings had multiple resales over many years resulting in up to seven resales for some works of art. Each resale pair was considered a unique point in our database. If the work of art was sold overseas, we converted the sale price into US dollars using the long-term exchange rate data provided by Global

Financial Data. Our data has continuous observations since 1871 and has sufficiently numerous observations to allow us to develop an annual All Art index for all paintings in the three categories combined from 1875 to the present (see Mei, Moses 2001). For this paper we have separated the data into three popular collecting categories and will use only data from the 1950-2001-time period. The first category is American Paintings (“American”) principally created between 1700 and 1950. The second category is Impressionist and Modern Paintings (“Impressionists”) principally created between the third quarters of the 19th and 20th century. The third category is Old Master and 19th century paintings (“Old Masters”) principally created after the 12th century and before the third quarter of the 19th century. For convenience, we call the first price from each price pair the “purchase price” and the second price the “sale price” adopting the perspective of the collector for

the time period between the two transactions corresponding to the price pair. The repeat-sales regression (RSR) uses the purchase and sale price of individual properties to estimate the fluctuations in value of an average or representative asset over a particular time period. Anderson (1974), Goetzmann (1993), and Pesando (1993) apply RSR to the art market. The benefit of using the RSR is that the resulting index is 4 Our data does not include "bought-in" paintings that did not sell due to the fact that the 4 Source: http://www.doksinet based upon price relatives of the same painting, which controls for the differing quality of the assets. Thus, it does not suffer from the arbitrary specifications of a hedonic model. The drawback of RSR is that the index is constructed from multiple sales, which are a subset of the available transactions. Olivier Chanel, Louis-Andre Gerard-Varet, and Victor Ginsburgh (1996) provide a detailed discussion on this weakness of RSR models.

We begin by assuming that the continuously compounded return for a certain asset i in period t, r i,t , may be represented by µ t , the continuously compounded return of a price index of art, and an error term: ri,t = µ t + ηit (1) where µ t , may be thought of as the average return in period t of paintings in the portfolio. We use sales data for individual paintings to estimate the index µ over some interval t = 1. T Here, µ is a T-dimensional vector whose individual elements are µ t The observed data consist of purchase and sales price pairs, P i,b , and P i,s , of the individual paintings comprising the index, as well as the dates of purchase and sale, which we will designate with b i , and s i . Thus, the logged price relative for asset i, held between its purchase date b i and its sales date, s i , may be expressed as  P ri = ln  i , s  P  i ,b = si ∑   =   µt + t = bi +1 si ∑r i ,t t = bi +1 si ∑ η i ,t t = bi +1 highest

bid was below the reservation price. 5 (2) Source: http://www.doksinet Let r represent the N-dimensional vector of logged price relatives for N repeated sales observations. Goetzmann (1992) shows that a generalized least-squares regression of the form µˆ = (Χ Ω −1Χ )−1Χ Ω −1r (3) provides the maximum-likelihood estimate of µ, where X is an NxT matrix, which has a row of dummy variables for each asset in the sample and a column for each holding interval. Ω is a weighting matrix, whose weights could be set as the times between sales as in Goetzmann (1993) or could be based on error estimates from a three-stage estimation procedure used by Karl E. Case and Robert J Shiller (1987) 3. Risk and Return Characteristics of the Art Price Indexes Figure One provides a graphic plot of the art indexes over the 1950-2001 period with the base year index set to be 1. Our reported art indexes are based on the three-stageleast-square procedure proposed by Case and Shiller

(1987) 5 The Adjusted R-squared for each of the three indexes is 0.71, 076, 060, respectively, suggesting that the art index 5 We use the Case and Shiller (1987) procedure because it allows us to adjust for a downward bias in annual returns estimation due the log price transformation (see Goetzmann (1992)). We have also estimated the art index using GLS and the two-stage Baysian estimation proposed by Goetzmann (1992). The correlation between the Case and Shiller (1987) procedure and the other two procedures are 0.970 and 0917, 6 Source: http://www.doksinet explains 71%, 76%, 60% of the variance of sample return variation for each category of paintings. The F-statistic equals 3927, 10125, 547, respectively, with a significance level equal to 0.000 for all three indices, indicating the indexes are a highly significant common return component of our art portfolios. Figure 1 shows a continuous rise in prices through the late1980’s when the art market peaked. The bubble burst first

in American paintings, with prices dropping 21% between 1989 and 1990. Impressionist painting prices followed, falling 51% between 1990 and 1991. Old Master painting prices fell 16% between 1990 and 1991. We can also see that, even after ten years of market adjustment, the Impressionist Painting Index has still not recovered from its high level in 1990. However, by 2000 the American and Old Masters indexes substantially surpass their 1990 values. Thus total performance of art during 1950-2001 period is affected by the bear market in art between 1989-1995 just as the total performance of the NASDAQ will be affected by its’ precipitous rise and fall over the last ten years. While the boom and bust was well documented in the art market, the price indices allow us to estimate the precise time and magnitude of the price change. Our indexes also identify major price declines during the 1974-75 oil crisis as well as during other recessionary periods. In addition, the indexes show that the

art price declines produced by recessions tended to be short term, were generally not seen until the second year of the recession and were followed by robust recoveries. suggesting that the results are quite robust. We have also discovered that the two-stage Baysian estimates tend to have smaller estimation errors though they may be biased. 7 Source: http://www.doksinet Figure 1. Mei/Moses Art Collecting Category Indexes 1000 American Impressionist Old Master 100 10 1 0.1 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 Table One provides summary statistics on the behavior of returns for the S&P500 and the three art indexes. The returns are all computed in nominal values For each variable, we report the mean, standard deviation, and its correlation with other assets. Table One reveals that for the 1950-2000 period American art had an annual compounded return of 12.2% comparable to the S&P 500, which gained 126% Moreover, we found that the volatility

of art market price indexes for the three categories were 42.9%, 335%, 301% during the 1950-2000 period compared to 15% for the S&P 500 index, making the art indexes more risky than the stock index. This is to be expected since the S&P 500 is a broad multi-industry based index while our individual collecting categories are quite narrow. In our previous paper we report on an All Art index, which 8 Source: http://www.doksinet has volatility closer to that of the S&P 500. Because of lower correlation with other assets reported in Table One, our study suggests that a diversified portfolio of artworks may play a role in portfolio diversification. Table One - Summary Statistics of Nominal Returns (1950-2000) American Impressionist Old Master S&P 500 Mean .122 .116 .112 .126 S.D .429 .335 .301 .150 American 1.00 Impressionist -.195 1.00 Old Master .007 .196 1.00 S&P500 .248 .232 .043 1.00 4. Art and Stocks during Wars and Recessions In

the fall of 2001, the $500 million art auction market was full of uncertainty. After the terrorist attacks on September 11 and the start of fighting in Afghanistan, art collectors became concerned about the impact of a slowing economy and a drawn out period of armed conflict on the world art market. By looking at the movement of art prices during wars and economic recessions of the past, we can draw some conclusions about how art prices may behave in future periods of war and recession. Using Business Cycles dates from the National Bureau of Economic Research and the Mei/Moses All Art Index, we have found several interesting price patterns. First, there tends to be a flight of funds to safety away from the art market during recessions, 9 Source: http://www.doksinet but the declines in art prices tend to be short-term and the average art price decline was only 0.7% during the 27 recessions over the 1875-2001 period Several recessions caused large price declines in the All Art index

as well as in the individual collecting category indexes, such as those in 1973-75 and 1990-91. However the declines in art prices did not become apparent until the second year of the recession and they were followed by robust recovery after the recession. Second, art prices are unpredictable even in the midst of a recession. While art prices generally tend to fall during a recession, this is not always the case. For example, there were no decreases in the All Art price index in the recessions of 1960-61, 1980, and 1981-82. In 1960-61 the American and Impressionist indexes were flat but the Old Masters index increased dramatically. In the 1980 economic downturn all three indexes increased. During the 1981-82 recession the American and Old Masters indexes were flat but the Impressionist index increased by about 10%. During the armed conflicts of lengthy duration of the last century, the art indexes out-performed major stock indexes. Using the data available from Global Financial Data

(http://ww.globalfindatacom) and the Mei/Moses Art Index, (http://www.meimosesfineartindexorg) the following comparison can be made During the extended World War One period, between 1913 and 1920, London stock prices as reported by the Financial Times All Shares Index reached a wartime low in 1918, having lost about 25% of its 1913 value over the period. After 1918 the index rose slowly to reach only 94% of the 1913 value by the start of 1920. In the United States the S&P 500 reached a wartime low in 1918, having lost 26% of its 1913 value. The S&P 500 then rose to 94% of its 1913 value by the start of 1920. In contrast, the Mei/Moses All Art Index declined 34% between 1913 and 1915 then rose to 125% of its 1913 value at the start of 1920. 10 Source: http://www.doksinet During the extended World War Two period, from 1937 to 1946, London stock prices as given by the Financial Times All Shares Index declined steadily from 1937 until mid-1940, losing about 50% of its

1937 value. It then rose and exceeded the 1937 value in mid-1944 and continued to rise to 107% of its 1937 value by 1946. In the United States, the S&P 500 declined rapidly from its 1937 value and had fallen almost 50% by early 1938. It reached its war time low by the spring of 1942 It then rose steadily and slightly passed its 1937 value in early 1946. In contrast, the Mei/Moses All Art Index increased from its 1937 value by almost 88% by the end of 1939, and finished at 130% of its 1937 value by the start of 1946. During the Korean War period (1949-1954) the S&P 500 grew steadily so that its value at the start of 1954 had increased 67% over its 1949 value. The Mei/Moses All Art Index also increased during this period. At the start of 1954 it was 108% higher than its 1949 value. At the start of 1954 the American Painting Index had increased by only 13%, Impressionist Painting Index had increased by 167% and the Old Master Painting Index had increased by 109% over their 1949

values. During the extended Vietnam War period (1966-1975) the S&P 500 cycled around its initial 1966 value. It reached a wartime low in 1975 having fallen by 27% of its 1966 value. The Mei/Moses All Art Index however grows almost continuously throughout this period with two intermediate declines of 20% in 1970 and 1974. The index level in 1975 is 256% higher than its value in 1966. In 1975 the American Painting Index is 244% higher, the Impressionist Index is 240% higher and the Old Master Index is 209% higher than they were in 1966. Thus in the two most recent periods of extended armed conflict, all four Art Indexes have out performed stocks except for American Paintings between 1949-1954. This is not a total surprise since American and Impressionist 11 Source: http://www.doksinet Paintings have outperformed the S&P for most of the second half of the 20th century (see the time period 1960-1996 in Figure 2). COMPARISON OF STOCK AND ART PRICE INDEXES DURING WARS World

War I (1913-1920) FTASI S&P500 MMAAI 1913 100 100 100 World War II (1937-1946) 1918 75 74 250 1920 94 94 125 FTASI S&P500 MMAAI Korean War (1949-1954) S&P500 MMAAI MMAMI MMIMI MMOMI 1949 100 100 100 100 100 1937 1938-1940 100 50 100 50 100 188 1946 107 101 130 Vietnam (1966-1975) 1954 167 208 113 267 209 S&P500 MMAAI MMAMI MMIMI MMOMI 1966 100 100 100 100 100 [EACH BASE YEAR (1913,1937,1949,1966) EQUALS 100; FTASI MEANS FINANCIAL TIMES ALL STOCK INDEX; MMAAI MEANS MEI/MOSES ALL ART INDEX; MMAMI MEANS MEI/MOSES AMERICAN ART INDEX ETC.] 12 1970 73 171 170 189 139 1975 73 356 344 340 309 Source: http://www.doksinet Figure 2. S&P 500 vs Mei/Moses Art Collecting Category Indexes 1000 S&P 500 1950=1 American Impressionist Old Master 100 10 1 0.1 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 The above price patterns may give some guidance to art investors. First, it might be unproductive to attempt to time the art market

precisely, since one can hardly predict wars and the onset and duration of business cycles. Second, while art prices tend to be quite volatile during wars and recessions, art also tends to be a good long-term investment. Third, if art prices fall substantially during a recession, these declines may present bargain-hunting opportunities for the long-term collector. Fourth, art provides good diversification benefits due to the fact that it is not perfectly correlated with stock market movement during periods of market stress. 13 Source: http://www.doksinet 5. Oil Prices, US Inflation, the World Financial Markets and Changes in the Price of Art We were curious to see whether changes in the general level of prices affected art prices or whether changes in the price of some global commodity would affect the Art Price Indexes. We chose changes in US inflation levels as an indicator of general price changes. We chose oil as the commodity to be studied due to its geographically diverse

set of production locations and oil plays a fundamental role in the U.S economy In the past, several of the oil market shocks have produced U.S recessions We were also interested in finding out whether there was any relationship between changes in global equity wealth and changes in the Art Price Indexes in the United States. Our study assumes that changes in the global wealth are indicated by changes in the value of the S&P 500 in the U.S, the FTX 100 in Great Britain and the Nikkei 225 in Japan. Our analysis is based on simply regressing the annual change in the level of the Art Index for each individual collecting category on a constant; the change in level of the independent variable, such as oil prices in the U.S; and the same independent variable lagged by one year. For the regressions involving indicators of world financial markets, we also lagged the independent variable two years to test the possibility that only a longer-term wealth change would affect the art markets.

The regression results for oil are given in Table Two. The results for US inflation are given in Table Three and the results for changes in world financial markets are given in Tables Four and Five. Tables Two and Three clearly show that the percentage change in the price of oil, the percentage change in U.S inflation or the same variable lagged by one year, has any significant relationship with the change in the value of the Price Indexes for American, 14 Source: http://www.doksinet Impressionist or Old Master Paintings. None of the regressions has an R-Squared higher than 0.07 Table Four summarizes the results of regressing changes in the level of each of our three individual collecting category indexes on changes and the one year and two year lags in the values of the worlds major stock indexes, plus a constant. Out of the nine possible regressions only a few had statistically and economically meaningful results. Table Four shows that the Old Master Paintings Index is effected

only by the Nikkei 225. The R-Squared of the equation is 0147 and the constant term and the one year lagged change of the Nikkei 225 variable are positive and have significant t statistic at the p value of 0.03 or better We interpret this result to mean that there is a small demand for Old Master paintings in Japan and that some of the paintings are supplied from the U.S market Lastly, Table Four depicts a substantial interaction between each of the S&P 500 and the Nikkei225 with the Impressionist Art Index. The model using the S&P 500 variables has an R-Squared of 0.221 with a positive coefficient of the variable for percentage change lagged one year and a significant t statistic at the p value level of 0.002 The model using the Nikkei 225 variables has an R-Squared of 0297 with positive coefficients for both the percentage change and the one year lag percentage change variables and significant t statistics at the p value levels of 0.008 and 0011 respectively The finding

that changes in the value of the S&P 500 plays a role in Impressionist Index is not surprising. Impressionist paintings sell for an average price that is ten times more than the average price of paintings in the other two collecting categories. Some auctions result in an average price for Impressionist paintings in excess of one million dollars. These extremely high prices might very easily make them subject to a change in wealth effect. The same interpretation applies to the effect of changes in the Nikkei Index. Many Asian collectors, especially those from Japan, have been major 15 Source: http://www.doksinet purchasers of Impressionist and Modern Art and New York has been the main market for these paintings in the last half of the 20th century. It is well known that during the expansion of the Japanese economy during the 1970’s and 1980’s, Japanese collectors were avid purchasers of paintings in this category. Thus again given high average prices, it is not unexpected

that changes in wealth in Japan would affect art price levels. The regression on the Nikkei Index is the only case where both the most recent percentage change in value of an index and that value lagged one year are both statistically significant. The implication is that a change in wealth patterns in Japan over a two year period is required to affect the change in the Impressionist Index. The correlation between the percentage change in the Nikkei 225 and its one-year lag is only 0.17 and thus it is not implausible for both to be statistically significant in the same model. Since the Durbin Watson test value for this model is 2.8, the stated t statistics would be a lower of bound on the true value of the t-statistic. A plot of the auto correlation of the residuals also shows that auto correlation was not a cause for concern in this model. The previous discussion has looked at the effect of wealth changes in the three major consumption markets, U.S, Asia and Europe, on an individual

basis with the New York auction market. Clearly the combined wealth effects in these three areas should also have some bearing on the health of the auction art market in New York. Table Five summarizes the results of this analysis. We did not include the second year lag effects in these models since they never were significant in any of the previous models and we wanted to keep our degrees of freedom as robust as possible. Again the only really meaningful finding is for the Impressionist Index. Changes in this index seem to be well and reasonably explained by changes in both the S&P 500 and the Nikkei 225. The R-squared of the equation is 043 with an F statistic of 547 with significance level equal to 0.0003 Thus the total equation is highly significant The coefficients for the change and one year lag change in the Nikkei 225, as well as the 16 Source: http://www.doksinet change in the S&P 500, are all positive and have a significant t statistic at the p value levels of

0.008, 0033, 0015, respectively There is also little correlation between these variables. (between -2 to +2) Thus due to the expensive nature of these paintings and the fact that the major clients for these works over the last 50 years have been increasingly from Asia and the U.S is not surprising that changes in their combined wealth would directly affect changes in the Impressionist Index. It is well known that the rise and peak in the Nikkei 225 coincides with the rise and the peak of the Impressionist Index in the 1980-1990 period. To insure that it was not this effect that was solely responsible for our results, we re-ran our model splitting the time periods at 1980. The post-1980 results had an inferior F statistic value and lower Rsquared than the pre-1980 data Only one statistically significant variable, the one-year lagged change in the Nikkei 225 was found and it was in the pre-1980 model. Since the post-1980 regression equation is statistically weaker than the original

equation it cannot be argued that the total effect in the overall model was caused by the activities of the 1980-90 time period. The Old Master market does not appear to be influenced by any of the three stock exchanges. The American Art Market is influenced by changes in the S&P 500 and the FTS 100 lagged one year. Both have significant t statistics at the p values of 002 and 0.000 respectively However the S&P 500 coefficient is negative while the FTS coefficient is positive. The R-Squared for the combined model is 030 and the F statistic of 3.22 has a significance level of 001 The negative value for the S&P implies that one year after the S&P 500 goes down, the American Art market prices go up and vice versa, while simultaneously the American Art prices move in the same direction as the FTS. This makes little economic sense. It is more likely that due to the high correlation between the S&P 500 and the FTS (0.54), the model has over-weighted the FTS and 17

Source: http://www.doksinet counter-balanced that with the negatively weighted S&P. We would argue that these findings are due more to correlation between the independent variables than to causation. We also experimented with using interaction terms between the percent change and the lag variables, as well as using indicator variables that were based on changes in direction rather than the extent of the change. Each of these models yielded either no significant results or results that were weaker than the results already reported. Thus, as one would expect, it is the size of the change in wealth rather than simply its direction that has the greatest influence on art prices In summary, our results seem to confirm the anecdotal evidence that the art market usually lags changes in economic activity by about a year. It is not surprising that this result is focused in the Impressionist market since paintings in this market have an average auction price which is often 10 times the value

for other works of art. This should make them potentially more susceptible to any change in the wealth effect. 6 Table Two - Art and Oil (1950-2001) Summary Statistics American R-squared .0003 Impressionist .0498 Old Master .0476 Constant .109 [1.71] .126 [2.55] .109 [2.44] % Change in oil price .000024 [-.001] -.0124 [-.065] .207 [1.213] % Change in oil price-1yr lag -.0305 [-.105] -.343 [-.1524] -.136 [-.670] Note: [t-stat] 6 We lastly experimented with a shorter time period (1960-2001) to test whether a different starting year could substantially affect our results. We did not encounter a result that significantly changed the explanatory power of the models, or eliminated or changed the sign of any of our significant variables. Thus our results are quite robust Revised versions of Tables Four and Table Five for this shorter time period are available upon request. 18 Source: http://www.doksinet Table Three - Art and US Inflation (1950 – 2001) Summary

Statistics American Impressionist Old Master R-squared .002 .040 .0716 Constant .240 .274 .197 [1.86] [2.73] [2.30] -.671 1.09 .4152 [-.18] [.37] [1.67] -.099 -3.42 -4.70 [-.03] [-1.17] [-1.89] % Change in inflation % Change in inflation-1yr lag Note: [t-stat] 19 Source: http://www.doksinet Table Four – Art and Global Financial Markets - Individual (1950-2001) Summary Statistics American Impressionist S&P500 R-squared .053 .281 Constant .191 .114 [1.55] [1.30] Old Master .0167 .157 [1.81] % Change in S&P500 .643 [1.33] -.228 [-.66] .073 [.22] % Change in S&P500-1yr lag -.161 [-.34] .114 [3.35] .227 [.68] % Change in S&P500 – 2yr lag -.210 [-.42] -.199 [-.56] -.149 [-.43] .066 .297 .147 Constant .143 [1.49] .071 [1.08] .133 [2.11] % Change in NIK225 .300 [1.08] .529 [2.79] .123 [.68] % Change in NIK 225 – 1 yr lag -.205 [-.76] .488 [2.66] .401 [2.27] % Change in NIK225 – 2 yr lag. .444 [1.61 -.104

[-.55] -.228 [-1.26] .18 .122 .019 Constant .089 [.88] .188 [2.26] .120 [1.56] % Change in FTS100 .300 [1.06] -.323 [-1.39] .141 [.66] % Change in FTS100 – 1 yr lag .858 [2.95] .365 [1.53] .115 [.53] % Change in FTS100 – 2 yr lag -.097 [-.34] -.070 [-.30] .175 [.81] NIK225 R-squared FTS100 R – squared Note: [t-stat] 20 Source: http://www.doksinet Table Five - Art and Global Financial Markets-Combined (1950-2001) Summary Statistics American Impressionist Old Masters R-squared .305 .427 .139 Constant .114 .007 .059 [1.36] [.11] [.79] .501 .088 .186 [.98] [.24] [.48] Change in S&P 500 1 yr -1.18 .892 .221 lag [-2.42] [2.53] [.59] Change in FTS100 .115 -.316 .092 [.37] [-1.39] [.38] Change in FTS100 1.24 -.051 -.097 1 yr lag [3.96] [-.23] [-.41] Change in NIK225 -.056 .484 .143 [-.23] [2.78] [.77] Change in NIK225 -.013 .391 .410 1 yr lag [-.05] [2.20] [2.17] Change in S&P 500 Note:

[t-stat] 21 Source: http://www.doksinet 6. Conclusions This paper uses a new data set consisting of repeated sales of paintings and estimates an annual index of art prices for three collecting categories for the period 1950-2001. Based on this new data set, our study made several discoveries. First we find that all individual art-collecting categories have been more productive investments than all fixed income securities. The Impressionist and Old Masters Indexes tend to under perform the S&P 500 while the American Painting Index was on a par with the S&P 500 performance over the period. The three art indexes have lower correlation with other financial assets than has been found in previous studies. As a result, a diversified portfolio of artworks may play a somewhat more important role in overall portfolio diversification than had been previously recognized. Second, our study finds strong evidence of superior performance of all art collecting categories during recessions

and periods of long armed conflict. 22 Source: http://www.doksinet Thirdly, with respect to changes in art prices, we find little evidence to support the assertion that a change in price of oil or US inflation is related to a change in the value of our individual Art Indexes over the last 51 years. We do, however, find a statistically significant relationship between changes in wealth, as indicated by changes in the S&P 500 and Nikkei 225, and the Impressionist Index over the last fifty years. In addition, the high return and lack of influence of world financial markets on the American Painting Index indicate that investors interested in balancing an art portfolio should consider paintings from this collecting category. We would like to note, however, that our results might only serve as a benchmark for those artworks bought and sold at major auction houses. Our return estimates also could be biased due to sample selection. In addition, art may be appropriate only for long-term

investment enabling its high transaction costs to be spread over many years. 23 Source: http://www.doksinet References [1] Anderson, Robert C. (March 1974), “Paintings as Investment” Economic Inquiry, 12:13-26. [2] Baumol, William (May 1986), “Unnatural Value: or Art Investment as a Floating Crap Game,” American Economic Review, (Papers and Proceedings), 76: 10-14. [3] Buelens, Nathalie and Ginsburgh, Victor (1993), “Revisiting Baumol’s ‘Art as a Floating Crap Game,” European Economic Review, 1351-1371. [4] Case, Karl E. and Shiller, Robert J (September – October 1987), “Prices of Single-Family Homes Since 1970: New Indexes for Four Cities,” New England Economic Review, 45-56. 24 Source: http://www.doksinet [5] Chanel, Olivier, Gerard-Varet, Louis-Andre, and Ginsburgh, Victor, (1996), “The Relevance of Hedonic Price Indices”, Journal of Cultural Economics 20, 1-24. [6] Goetzmann, William N. (December 1993), “Accounting for Taste: Art

and the Financial Markets over Three Centuries,” American Economic Review, 83: 1370-6. [7] Goetzmann, William N. (March 1992), “The Accuracy of Real Estate Indices: Repeat Sale Estimators,” Journal of Real Estate Finance and Economics, 5: 5-53. [8] Mayer, Enrique, International Auction Records, New York: Mayer & Archer Fields, various years. [9] Mei, Jianping and Moses, Michael (September 2001), “Art as an Investment and the underperformance of Master-pieces,” American Economic Review (Forthcoming) 25 Source: http://www.doksinet [10] Pesando, James E. (December 1993), “Art as an Investment: The Market for Modern Prints” American Economic Review, 83:1075-1089. [11] Reitlinger, Gerald (1971), The Economics of Taste, Vol. 1, London: Barrie and Rockcliff, 1961; Vol. 2, 1963; Vol 3 26