What is the recent behaviour of the demand for art and how does it correspond to the demand
for financial stocks?
Agata Kik
Abstract
Recently, popularity of investing in art has sharply increased. Art started to be treated as a
serious financial investment. The previous research of the art market is divided between those
that claim that art can be a good way for portfolio diversification and those in whose opinion
it cannot be used for hedge activities. The answer to this problem is determined by the fact if
art and other financial assets are correlated or not. However, past empirical results are
disagreeing on this issue. This research investigates into demand trends in the art market in
the time period between 2007 and 2013 and examines their relation to the trends of the
demand in the financial market. From examination of trends in these two markets, it appears
that demands for art and financial stocks react to economic shocks in a similar way. The
decreased spending in the financial market, during the 2008 economic crisis, is not
accompanied by increased money transfers to the art market. Trends in art and financial
markets closely resemble each other, however any causal relationship cannot be proposed.
They appear to be independent, but exhibit similar tendencies of investment. The pattern of
the demand in the art market reacts to the analysed economic shock with a delay, when
compared with the financial market. The delay is explained not by financial market’s
influence on the art market but by lower degree of liquidity of art assets. Lower liquidity
causes them to react to the 2008 crisis with a time lag. This research provides some valuable
insights for investors, who are willing to put money into art and raises awareness of art
market’s performance being comparably vulnerable to economic shocks as the financial
market.
Structure
This paper consists of four sections. First of them introduces the topic of art as a financial
asset and shows results of the previous researches of the art market. The second section
presents my empirical collection of data and the method in which it is used. The third section
shows the obtained results based on graphical analysis and four regressions. The last section
discusses the theoretical framework and the empirical results of my investigation. It points out
some possible improvements and gives useful conclusions for investors interested in the art
1. Introduction and Literature
The art market has gained a widespread popularity in the recent part of the history.
Escalating prices and excessively high auction results are common news of the art market.
Collectors and art investors seem to be particularly encouraged to purchase art works.
Contemporary Art is the sector that has gained the most of investor’s interest in the recent
times. Even though the artworks of the latest decades involve the most of uncertainty
regarding their future value, their sales still reach the highest recorded bids in the auction
houses such as Sotheby’s or Christie’s. These two are among the oldest and the largest
auction houses in the world. Their main function is to intermediate transactions in the
artworks. They are attractive for artists as they create a secondary market for art and break the
monopoly of private art dealers.
Auction’s environment poses a good source of information to study the art market.
Galleries’ or private dealers’ sales are not transparent and do not pose a significant part of the
art market. Being uncertain about the former value of an artwork or the price it was purchased
for, it is sensible to focus on volumes and values of art sales instead on their realised returns.
This research analyses the trends of art buying, by focusing on total spending on art in the
major auction house: Sotheby’s. An important thing to point out is the wide range that ‘art’
encompasses at this auction house. The objects auctioned include fine art such as paintings
and sculptures, jewellery, furniture, wine or manuscripts. In order to be objective about the
demand for any valuable artwork, this investigation looks at sales from all of these
departments. Moreover, this research takes into account the period between 2007 and 2013 in
order to look into the art market demand’s trends and its reaction to the financial crisis of
2008. Close attention is paid to the relation of the art market’s sales to the performance of
and preceding closure of many financial institutions. If the pattern of the demand for sales in
the art market is analogical to the demand for stocks in the financial market, the relationship
between them can be examined and their resemblance proposed. Even though art seems to be
far from being similar to any other financial assets, it does hold some of their characteristics
and many treat it as an attractive investment.
Only most recently art started to resemble financial stock’s characteristics of an asset.
For most of the decades of the twentieth century art was treated only on the basis of its
aesthetic values. However, its investment character was aptly predicted already in 1960s.
According to Leo Steinberg ‘Art is not, after all, what we thought it was; in the broadest sense
it is hard cash. (…) Another decade, and we shall have mutual funds based on securities in the
form of pictures held in bank vaults’ (Other Criteria, Lecture given at the museum of Modern
Art, New York, 1968). He was totally correct, in what he forecasted as nowadays we observe
art funds’ proliferation. The fact that art started to be treated as a serious financial investment
was closely related to increasing transparency of its market. Thanks to many art price
databases and indices, investors can make rational decisions based on objective statistics.
Popularity of art funds is increasing. According to Campbell (2008) they offer diversification
strategy possibilities, as art is characterised by low correlations with other assets, which
investors hold in their portfolios. Funds, which focus on investment in art, present us with the
situation when art is an investment for purely financial gains. To what Campbell (2008) draws
attention is the fact that there is a possibility of higher profits, when the resources are pooled
and the help of experts is available (p.78). It can be predicted that, while being in their
infancy, art funds await their dynamic growth.
The transparency of the art market is enhanced by construction of multiple art indices
that allow for comparisons with different markets. Candela and Scorcu (1997) presented a
allows one to have a complete view of the art market with a possibility of focusing on its
certain segments (p. 193). Their study introduced a new approach to art price index’s
construction. For the first time, instead of using an average painting method, which is
sensitive to the passage of time, the time-steady ‘representative painting’ was found. This
method was based on aggregating the frequencies and creating a distribution out of paintings’
market and estimated prices. According to Candela and Scorcu (1997) creating price indices
is necessary in order to identify the mechanisms of any market (p. 192). As my research wants
to analyse the trends rather than mechanisms, I assume the demands data will be sufficient.
Even though exact determination of causality or other relationships between art and financial
markets is not be possible, the correspondence of behaviour with its similarities and
differences can be observed. The previous research into the exact correspondence and casual
movements between assets of art and stocks poses a base for this study.
An interesting question concerns the factors that influence the buyers’ behaviour in the
art market. Art does not belong to a normal group of commodities, whose values can be easily
estimated or determined. Intrinsic aesthetics of art and additionally differing cultural tastes
cause the determination of the value of art to be so complicated and almost impossible. Works
of art are not very liquid and it makes them to appear very risky. Moreover, works of art are
not divisible and they involve high transaction costs and long delays before the actual sale
(Channel, 1995). Furthermore, there exist additional risks such as theft or forgery. However,
there should be no doubting if art can be called an asset. A work of art is clearly a valuable
object and it is a property that can be owned by a person or an institution. The value of art is
also inflated by the uniqueness of each piece. The qualities that determine its worth include:
the period of creation, attributions, presence at major exhibitions, an artist’s status, the subject
or formal features. All of these aspects contribute to the value of a work of art and therefore
affected by so many elements posed a rationale behind Renneboog and Spaenjers (2013)
investigation into prices and returns in the art market. Due to the fact that returns on art assets
cannot be compared straightforwardly, Renneboog and Speanjers (2013) decided to use a
method of hedonic regression. Hedonic regression meant including dummy variables, which
represent the characteristic that might have had an impact on the dependent variable. Research
into returns on art between the years 1957 and 2007 showed that factors such as artist
reputation, attribution, signs of authenticity, medium, size, topic, timing and location of the
sale were highly related to the price levels of artworks sold (Renneboog and Speanjers, 2013,
p. 36). Recent art studies use hedonic regressions more often in order to account for
ambiguous influences on price levels. However, as Renneboog and Speanjers (2013) pointed
out, the extent of difficulty of finding the right hedonic characteristics is vast (p. 38).
Most of the historical studies used simpler methods of regression. For example, Mei
and Moses (2002) used repeated-sales regression as the method to investigate performance of
art investment. The first thing they pointed out was the heterogeneity of artworks and
infrequency of their trade that caused the art market to be so difficult to research (Mei and
Moses, 2002, p. 1656). They tried to overcome this problem by simplifying their research to
repeated-sales data. They gathered data on works of art that were sold twice and then
compared their prices. They also constricted their method to only a few time periods of art
history. Pairs of prices enabled them to create art indices. The simple regression by Mei and
Moses (2002) did not take into account any additional factors influencing art prices. Its
examination of the ‘law of one price’ posed an extension to the previous research (p. 1656).
The ‘law of one price’ would imply that the place, where the artwork is sold, should not
influence its price. However, according to Pesando (1993), not only does the price of similar
prints vary at different auction houses, but also it differs between countries. Pesando (1993)
and they did about 14% higher at Sotheby’s compared to Christie’s (p. 1088). Nevertheless,
Mei and Moses (2002) proposed a different technique than Pesando (1993) and obtained
unclear evidence. Even if there existed some price differentials the differences between them
were very small and simply insignificant. In my research I assume that the same artwork
would achieve the same price regardless of the location it was offered at.
My representation of the financial market is based on the S&P 500 index. It is the
most common index of the financial market used in the previous studies. For example, Mei
and Moses (2002) made use of this index to proxy the systematic (affecting the entire market)
risk in the financial market. After having made regressions for art and financial markets, Mei
and Moses (2002) compared their betas. Betas demonstrate the sensitivity of the return on a
specific asset compared to the expected return on the market and therefore are indicators of
the market, un-diversifiable, risk. What Mei and Moses found was that art assets had smaller
betas than the financial assets (p. 1663). As a result art can be considered as involving less
systematic risk than financial stocks. Due to the existence of a trade-off between risk and
return, art was expected to have lower returns than stocks. On the other hand, when compared
with bonds art had a higher systematic risk and therefore should have been more profitable in
the long run (Mei and Moses, 2002, p. 1663). Accordingly, the demand for art can be
influenced by the fact that it outperforms the fixed income securities and involves less risk
than the stocks. Mei and Moses (2002) also proposed that art index is less volatile and less
correlated to other assets (p. 1666). Such characteristics mean that art, as an investment,
should be a good way for portfolio diversification. The lower the correlation between assets
the more benefits there are from diversification. Campbell (2008) as well examined art’s
performance in a diversified portfolio. His empirical research into optimal portfolio
allocations concluded that art could be an attractive contribution to an investor’s strategy (p.
Studies of the art market do not agree on the characteristics of art as an asset.
Contradictory to what has been just showed, some researchers claim that art cannot pose good
diversification options. For example, Goetzmann (1993) claimed that art couldn’t be used as a
hedge against financial stock’s fluctuations. In his research, the extreme risks that art
investments involved did not imply any extra higher returns (p. 1374). Goetzmann (1993) as
well researched the art market trends. Using the art returns between years 1720 and 1990 he
observed periods of bull and bear markets. Bull markets mean increased demands, increasing
prices and therefore exceptionally high returns. On the contrary, bear markets show periods,
when prices are falling and the pessimism dominates the demand side. Bear markets in
Goetzmann’s (1993) study respond to economic recessionary periods in Britain and the
United States (p. 1373). According to this, the last economic crisis, which was spread all
around the world, must have brought similar results. My research investigates if the financial
stocks’ crash in 2008 led to an increased investment in art due to hedge reasons, or maybe
contrary, it resulted into a bear market in the art market. While comparing means, standard
deviations and correlations between art and stock returns, Goetzmann (1993) got to the
opposite result than the research by Mei and Moses (2002). His examination of returns led
him to conclude that art assets are highly volatile and strongly positively correlated to other
assets and especially to stocks (p. 1374). Even though art returns appeared to be high, the
above-mentioned characteristics prevented it from being perceived by investors as highly
attractive. It is only the risk-neutral investors, who would treat art as an excellent investment
and according to Goetzmann (1993) the high correlation between the art, the stock and the
bond markets makes it a poor vehicle for portfolio diversification (p.1375). As more recent
studies find contrasting results, my research will try to inquire into the link between art and
One of the reasons that investment in art is growing so rapidly is its aesthetic dividend
that sole cash dividends of a stock are not able to replace it. Goetzmann (1993) concluded that
as the objective aesthetic valuation of an artwork is nearly impossible, it is the wealth of the
collectors that determines its price. On a free market forces of the demand and supply
determine prices, but in the art market prices are determined by the wealth of art buyers. This
is what makes art so closely related to stocks. As the value of stocks on the financial market
determines the wealth of an investor, the value of stocks will directly influence his demand
for art. The buyer will choose art from other similarly volatile assets, as it supplies him with
extra aesthetic profits. In order to investigate the relation between stocks and art, Goetzmann
(1993) performed a Granger test. The result showed a strong evidence of casual relation
flowing from the stock market to the art market. The demand on the art market followed the
demand on the financial market, as changes in art prices went along with changes in stock
prices. This is, according to Goetzmann (1993), an implication for the stock’s price effect on
wealth of investors and further influence on the demand in the art market. His result was
proved in the study of Campbell (2005). Her suggestion to rather base the forecasts for the art
market on lagged variables of both stocks and art is used in this paper. I expect a delayed
change of the demand in the art market, which responds to the instant change in the demand
for stocks, coinciding with the 2008 crash.
The time, I have chosen for this research, is the period between 2007 and 2013. These
years include the time of the 2008 Global Financial Crisis. This crisis is often claimed to be
the worst one since the Great Depression of the 1930s. Many major financial institutions
failed and many banks had to be bailed out by governments. The 2008 crisis resulted in a
major downturn of financial stocks, what meant that they prices dramatically fell down. In
November 2008 S&P 500 index went down by 45% from its 2007 high. Statistics show that
worth. Clearly this huge drop of stocks’ prices meant that investors from around the world felt
less wealthy. As the value of their financial assets extremely diminished and the financial
market crashed to a great extent, investors might have turned their attention to other
non-financial assets, one of them being art. This paper investigates the reaction to the non-financial
crisis as a shock in the art market. There are two possibilities of reacting to the shock: moving
funds from one group of assets to the other one, or decreasing spending in each asset’s
market. What I predict to be more probable is the ‘wealth effect’ that Goetzmann (1993)
presented. Moreover, according to the Granger test’s results in previous studies, art follows
stocks’ behaviour. While during the 2008 financial crisis, demand for stocks fell, I expect a
retarded fell in the demand in the art market. A lagged behaviour of the demand for art was
also explained in the investigation of art, financial and real estate markets by Candela and
Scorcu (1997). However, their Granger tests showed that prices in art and financial markets
were independent of each other. Candela and Scorcu (1997) presented an interesting finding
of the real estate market being related to the art market (p. 192). According to them, it is the
art market that predicts the real estate market and this results from the fact that the first is
more liquid than the other.
The scepticism, about buying artworks, results from some obvious reasons. Firstly,
even though art indices predict quite good long-term returns, they are not able to show with
certainty the artist’s names or art periods, which will be the most profitable in the future.
Tastes and fashion are hardly ever foreseeable. Investment risks and ambiguity of art
discourage investors, preventing them from taking into account any diversification benefits
that might result in the future. Chanel (1995) admitted that art might be treated as a
speculative asset (p. 520). His research tried to investigate the predictability of the art market.
As Goetzmann (1993), he also found that the relationship between financial and art market
relationship between these two markets and they tended to move together as no systematic
transfers took place between them (p. 527). His research gave an idea about the fact that the
demand’s behaviour in the art market might have been predicted by stock exchanges.
Financial markets immediately react to economic shocks such as the 2008 crisis, while art
markets respond with a delay. Chanel (1995) claimed that as the profits earned in the financial
market could be invested in art, performance of trade in stocks could be a great guide for the
art market’s predictions. Nevertheless, the aesthetic gains, for which Frey and Eichenberger
(1995) accounted as ‘psychic benefits’, leave this issue ambiguous. This implies that only the
art collectors, who value their cultural enrichment from investing in art, would be the
potential art buyers. Still, art funds are on the rise and their popularity is spreading among
investors of purely financial reasons. As the popularity of art as a financial asset is increasing,
the investigation into its market poses a great source of knowledge for investor’s future
strategies and actions. Especially the fact that this research is based on the most recent time
period between 2007 and 2013, brings an invaluable extension to the previous examination.
Moreover, the 2008 Economic Crisis is looked at from the perspectives of both art and
financial markets. This gives a better understanding of trends of the markets that coexist in the
current global economy, which is so intertwined and unintelligibly advanced. What is more, it
cannot be argued anymore that researching such a non-transparent market, as the art market,
does not bring concrete knowledge or does not pose enhancement to the economic theory. No
longer is the art market characterised by confidentiality and the transfers that take place in it
2. Data and Methodology
For the art market research, the data I collect are taken from one of the two major auction
houses in the world: Sotheby’s. It is a renowned institution, which was founded in London
with its headquarters currently based in New York. It is one of the two biggest art brokers
rivals, competing in its sales with the second most famous auction house: Christie’s. It
specializes in sales of collectible pieces, of which fine art constitutes the majority. Even
though it does not represent a whole market for art, it can clearly show the patterns of trade
that occur inside this market. Being one of the most internationally known auction houses, it
is the first institution that a potential art buyer or seller would turn to, in order to make a
transaction. It can be imagined that the art market is divided into sales in auction houses and
commercial art galleries. It is nearly impossible to account for each commercial art gallery in
the world. Each gallery’s sales pose insignificant numbers, when compared to the total trade.
Moreover their dispersion introduces geographical and cultural differences, which are not
easily accounted for. As Sotheby’s and Christie’s are in severe competition and they
approximately equally share the current auction market for art, treating one of them, as an
example for the entire auction house’s market, is justifiable. According to some statistics,
auction houses include approximately half of the whole trade in art and therefore stand for the
most important transactions that happen. As volumes and values of auction houses’ trade are
so substantial, changes in its demand and supply must be highly consequential for the rest of
its market. Because this paper aims to analyse trends rather than real values, the choice of a
sample from an auction houses’ art market further their finest representative – Sotheby’s
should be agreeable. This work examines trends and it thus looks at the behaviour of the
Volume of sales is the criterion used for the art market’s sample. The Sotheby’s online
database serves as a source. Auction results are collected and systematically organised. Time
series are created out of the total values of sales executed during each auction that has
happened between years 2007 and 2013. Each auction’s total sales are expressed in US
Dollars for the possibility of worldwide comparisons. The data is browsed on the 7th and 8th of December 2013. The data on the auctions that have happened between the 8th and 31st of December is browsed on the 6th of January 2014. The exchange rates are used as of these dates. The locations of auctions that have happened include: Amsterdam, Beijing, Geneva,
Hong Kong, London, Milan, New York, Paris, Toronto, and Zurich. Majority of auctions, in
similar proportions, is divided between London and New York. Even though the number of
cities, where auctions happened is limited, the data from Sotheby’s can be treated as a global
example. In fact, to take part in bidding does not require one’s physical presence at the
auction, so the constraint of the location is absent. In the end, quite a lot of major collectors
and art investors hire assistants at Sotheby’s to act as their brokers. What is more, it is quite
popular to make bids through phone calls. After each auction, the hammer prices of lots, for
which they have been sold, are summed and they are expressed as auction’s total sales. I
record total sales from each auction and then sum them up into total monthly sales. I do this
for a clearer representation of the data. There is of course a possibility of an offered lot not
being sold. Due to this fact the total numbers of each auction’s lots will not be taken into
account. The sample of 1945 auctions from the years 2007 to 2013 represents all of the
auctions that happened at Sotheby’s. All categories of collectible objects are included. The
majority is constituted by fine and decorative art auctions. Nevertheless, auctions of
jewellery, furniture, watches and wine are included. This is done in order to omit any
possibility of collectors switching from one category to the other. Inclusion of all the
different art classes. Moreover, the seasonality of the art trade needs to be kept in mind. The
numbers of auctions highly differ between months. For example, in the month of August in
the period between 2010 and 2013, there were no sales executed, while in some months the
number of auctions added up to almost 50. Due to these huge differences in sales between
different months I as well examine yearly statistics. I predict that yearly data will pose a better
graphical depiction of the effects of the 2008 crisis on art and financial markets. Nevertheless,
as financial market’s data are expressed as monthly time series, regressions are based on
monthly intervals.
My sample size for the art market amounts to 1945 recorded sales that happened between
2007 and 2013. The number of sales differs between months. The seasonality of trade in the
auction house is shown in Figure 1. The y-axis shows the number of auctions that were
organised in the given time period. The x-axis labels represents periods divided into year’s
quarters. The series starts with the first quarter of 2007 and ends with the fourth quarter of
2013. From this graph it is visible that high seasons of trade in art occur in spring and autumn,
while low seasons take place in winter and summer. The seasonality of activity in the art Figure 1 Number of auctions organised by Sotheby’s in each quarter between the years 2007 and 2013
market is important to notice, as it is not usual in other markets. Table 1 and Table 2 show the
total sales’ values and the corresponding numbers of auctions organised by Sotheby’s in the
period between 2007 and 2013.
Year Total Yearly Sales
2007 4750806047 2008 4525480615 2009 2277097817 2010 4452157232 2011 4874585222 2012 4443984744 2013 4976480703
Year Number of Auctions
2007 360 2008 308 2009 254 2010 279 2011 258 2012 235 2013 251
Table 1 Values of Total Yearly Sales in Sotheby’s between
2007 and 2013 expressed in US Dollars, with the average equal to 4 328 56 054,29 (USD).
Table 2 Values of Number of Auctions organised by Sotheby’s in each year
between 2007 and 2013, with the average of 277 auctions in a year.
Figure 2 and Figure 3 give a better view of the trends in the art trade that happened at the Sotheby’s auction house in the period between 2007 and 2013. Even though the number of
auctions did not differ significantly in this time period, there was a huge drop in the value of
total sales in 2009. Total sales between 2008 and 2009 decreased by more than 2 billion
(USD). This is a half of average total sales for this time frame.
Furthermore, the financial market is analysed on the basis of S&P 500 Index. It is a
stock’s price market index, which is based on the results from 500 largest companies. It is
used as a broad representation of the situation on the financial market. S&P 500 is used in this
paper as a general image of the price and demand trends in the market for stocks. The
recorded values are the Adjusted Closing Prices in each month between January 2007 and
December 2013. These are the most accurate variables used to analyse the historical
performance of stocks, as they show stocks market capitalisations. Their values include all Figure 3 Number of Auctions organised by Sotheby’s (shown on y-axis) in each year between 2007 and 2013 (x-axis).
corporate actions that has occurred in a given month. From YAHOO! FINANCE database
Adjusted Closing Prices for S&P 500 Index are collected. These are expressed in Figure 4.
Year Average year's Adjusted Close
2007 1 478,10 2008 1 215,22 2009 948,52 2010 1 130,68 2011 1 280,76 2012 1 386,51 2013 1 652,29
Figure 4 Adjusted Closing Prices for the S&P 500 market index, represented monthly between January 2007 and December 2013. y-axis depicts prices in US Dollars of the stock index for each month. December 2013 is characterised by the maximum value of stock- 1848,36 USD, while February 2009 shows the minimum value during these years- 735.09 USD.
Table 3 Computed from monthly statistics, the averages of each year’s Adjusted Closing Prices of S&P 500 for the years 2007 to 2013, expressed in US Dollars.
For the comparison with art market sales, the monthly results for S&P 500 are
transformed into yearly averages and summarised in Table 3. Table 3 is presented below in
form of a graph in Figure 5.
The time period for the time series of total art sales and stock prices is the years
between 2007 and 2013. These years are chosen in order to include the 2008 Financial Crisis
that is claimed to start with the end of 2007. Moreover, previous research of the art market
stops its analysis on 2007. This paper provides extending insights into the market’s trends,
based on the most recent data. Besides, the effects of the last crisis are observed and their
influence on investment patterns in both of the investigated markets is shown. Both groups of
variables used in this research, the S&P 500 price index and total sales Sotheby’s, are
expressed in US Dollars. Their values have been averaged to present time series of yearly
intervals. Sotheby’s total sales are treated as a representative for the art market, while stock
prices depict the situation on the financial market. Changes in the values of the total auction
sales express the behaviour of the demand and willingness to invest in the art market. On the Fig. 5 Each year’s averages for Adjusted Closing Prices (USD) of S&P 500 Index.
Equation 1 Lag relationship between the dependent variable with the current value and lagged values of the explanatory variable , where is a constant and is the error term.
other hand, financial market’s trends are analysed with the aid of stock prices, as the sole
demand side of trade in stocks cannot be traced. By treating the stock market as the case of a
perfect competition market model, where equilibria are determined by forces of the demand
and supply, changes in stock prices are as well a valid representation of changes of the
demands in the financial market. Changes of stock prices communicate changes of the
demand of financial investors. A decrease in a price of a stock shows a decrease in the
demand for it and respectively an increase in stock prices is associated with an increase in the
demand for it. As this paper investigates relationships between the art and financial markets,
the method used for the empirical research is the Polynomial Distributed Lags (PDLs). The
distributed lag model is often used in econometrics for time series data. With PDL the
estimated regression predicts the current values of the dependent variable based on both
current and lagged values of the exogenous variable. The endogenous variable in this model is
the total art sales. The predicted result for this study is that demands in the art and financial
markets are positively related. This means that stocks and art assets react to the shocks in the
economy in similar ways. However, changes in the behaviour of the investment in these two
markets are expected to occur with certain lags. The demand patterns for art are forecasted to
follow demand trends for stocks with certain delay.
There are four regressions carried out. The distributed lag regressions have the form of
The number of parameters to be estimated by PDL is reduced in order to impose a
smoothness condition on the lag coefficients. The coefficients have to lie on the
polynomial of a relatively low degree, as represented by Equation 2.
Polynomial of the 3rd degree is chosen. The polynomial of a low degree allows for capturing the true lag distribution. Due to the fact that the chosen lag length is large (12 lags as shown
in Equation 1), the multicollinearity might occur and therefore severe consequences on the
regression might result. Choosing a low-degree for the PDL allows us to avoid this problem.
The number of 12 lags is chosen in order to account for yearly effects, because the data is
divided into monthly values. The time period of a whole year allows also for bypassing
seasonality problems. Winter and summer total sales drop and such decreased values would
impose a negative bias on the regressed values.
Firstly, the total sales in Sotheby’s are regressed on their own lagged values. In order
for the regression to be as precise as possible the monthly data is used. There are 84
observations for each of the variables (12 months during 7 years). This research conjectures
that the past (lagged) values have influence on current values of the dependent variable.
Previous spending, with 12 lags capturing a year’s interval, is predicted to have influence on
the current one. This comes from the fact that previous spending, and therefore past demand,
expresses the earlier situation on the art market. In times of market’s downturns, for which
2008 economics crises might be an example, total art sales are expected to drop and such a
drop with spread pessimism among art buyers clearly affects their future/current spending.
This regression is needed, as when examining the demand for art sales, the influence of its
past behaviour on its current and future values needs to be examined first, in order to be able Equation 2 Coefficients lying on the 3rd degree polynomial of the distributed lag model.
Equation 3 Current monthly art sales at Sotheby’s ( ) are regressed on 12 lagged values of the previous monthly total sales at Sotheby’s. Total sales are expressed in US Dollars. is a constant and is the error term. This equation is based on 3rd degree polynomial lag distribution for .
Equation 4 Current monthly art sales at Sotheby’s ( ) are regressed on a current and 12 lagged values of the previous monthly adjacent closing prices of the stock index S&P 500 ( ). Total sales and stock prices are expressed in US Dollars. is a constant and is the error term. This equation is based on 3rd degree polynomial lag distribution for .
to examine other additional external influences for example stocks’ prices values. Equation 3
shows the regression of the current total Sotheby’s monthly sales dependent on its 12 lagged
values.
Next, the total art sales will be regressed on the current and past values of stock prices.
This regression needs to be performed in order to examine if stock’s past prices alone have
any significant effects on sales in the art market. The reasoning behind this regression is the
‘wealth effect’, the fact that the stock price expresses the value of an investor’s assets. If the
stock is expensive, the assets seem to have high value and investors feel rich. The impression
of increased wealth, results into higher spending on luxurious assets such as art. It is predicted
that when the prices on the financial market go up, the total sales in the art market increase
too. Equation 4 shows this regression.
The third regression will be based on lagged values of both the endogenous variable of
the total art sales and the exogenous variable of stock prices. This is performed in order to
increase the explanatory power of the second regression. This regression is necessary in order
to inspect if combining the explanatory variables of the two previous regressions improves the
preciseness of the predictions. It is probable that stock prices cannot exactly predict the values
Equation 5 Current monthly art sales at Sotheby’s ( ) are regressed on a current and 12 lagged values of the monthly adjacent closing prices of the stock index S&P 500 and 12 lagged values of total monthly sales at Sotheby’s. Total sales and stock prices are expressed in US Dollars. is a constant and is the error term. This equation is based on 3rd degree polynomial lag distribution for .
Equation 6 Difference of the current art sales with the previous month’s sales at Sotheby’s ( ) are regressed on a current and 12 lagged values of the monthly differences of adjacent closing prices of the stock index S&P 500 and 12 lagged values of the differences in total monthly sales at Sotheby’s. Total sales and stock prices are expressed in US Dollars. The sample size decreased to 83 observation for each variable due to consideration of differences. is a constant and is the error term. This equation is based on 3rd degree polynomial lag distribution for .
results are expected to poses a higher explanatory power. The comparison of significance will
be based on comparing the R2 values of the performed regressions. Such a model aims to catch the linear interdependencies of the time series of variables from both art and financial
markets. Equation 5 presents its form.
The last regressed model will take into account the differences’ dependencies. The
difference of total Sotheby’s sales between consequent months will be regressed on their
lagged values and on the lagged values of differences in stock prices in the responding
months. Such a regression will be a great extension to the previous models as it can trace the
direct effects of changes of the demands in art and financial markets. It is necessary for my
research as it incorporates the relative effects of changes. As this paper investigates the trends
and therefore changing behaviours of demands in the two markets, estimation of relationships
between changes of art sales and changes of stock prices is essential. Equation 6 presents the
last model of this empirical research.
3. Results
As this work looks at relationships and trends of the demands in the art and financial
S&P 500 is shown in Figure 6. The total yearly sales at Sotheby’s are expressed in US
Dollars. The price index taken from S&P 500 is scaled in order to be within the boundaries of
the art sales’ graph. Each of the averaged yearly adjusted closing prices from Table 3 is
multiplied by 10000000. I allow myself to scale the stock prices as I am researching trends
rather than real values. Scaling average stock prices by 10 millions allows for graphical
comparison with total art sales. In the graph there appears a striking similarity between the
two lines of the total Sotheby’s sales and scaled yearly average price indices of S&P 500.
Both the art sales and the stock prices reach their minimum in the year of 2009. Even though
the lag was expected, the minimum values are reached in the same year. The lag, however,
can be seen at the beginning of the analysed period. In particular, in 2007 the stock prices
already started to decrease rapidly. On the other hand, it was only in 2009 that Sotheby’s
experienced a huge drop in sales. The data represented in Figure 6 depicts an influence of an
economic downturn on markets for art and for stocks. These two markets react to the
Figure 6 Total yearly sales at Sotheby’s (USD) presented with scaled (10 000 000 larger) average year’s stock prices (USD) from S&P 500 Index between the years 2007 and 2013.
economic shock of the 2008 crisis in the same way. The economic crisis causes demands for
art and for financial assets to fall. It is clearly visible that when prices and so the demand for
financial stock decreases, investors do not turn to the art market. When the situation in the
financial market worsens, investors feel as they are loosing money and contract their
investment in both of the markets. My data embracing the last 7 years, which include the
time of the Global Financial Crisis, shows that changes in investment behaviour in the
financial market are followed by positively correlated changes in spending patterns in the art
market. Increased stock prices, which result in higher worth for financial investors, are
accompanied with increased spending in the art market.
The Polynomial Distributed Lags regression of total art sales on its lagged values
resulted in the statistics summarised in Table 4. The Time Lags 1-4 correspond to the
third-degree polynomial of the predicted values of 12 lagged values of the dependent variable of
total art sales. The Time Lag 1 corresponds to the current value of the art sales, the lag of
period 0. Time Lags are the values of the Equation 7, similarly presented by Equation 3. C is the constant ( ) included in the regression.
Variable Coefficient Std. Error t-Statistic Prob. C -96932058 69400963 -1.396696 0.1671 Time Lag 1 -0.170113 0.023116 -7.359167 0.0000 Time Lag 2 0.014167 0.021311 0.664753 0.5085 Time Lag 3 0.019151 0.001115 17.17053 0.0000 Time Lag 4 -0.001090 0.000765 -1.423475 0.1592
Table 4 Reported results for the regression Equation 7, where are the consequent time lags.
Equation 7 Estimated equation for the monthly Sotheby’s total sales, based on their lagged values, where are the values estimated on the basis of 12 lagged values of the Sotheby’s total sales and s are estimated values of the coefficients from Equation 3, is therefore the predicted value of from Equation 3, here presented in form of a 3rd degree polynomial. is a constant and is the error term.
The coefficients of the time lags are presented in the second column of Table 4.
In Table 4 it is important to notice the p-value of the coefficient of the first and third time
lags. P-value for these two coefficients is equal to 0 and it means that the term is unlikely to
happen and therefore should be rejected. At 5% significance level, the current value and the
second lag of art sales values can be avoided. They are not significant in explaining the
current total art sales. According to Table 5 R2 and F-statistic are very high. This means that previous lagged values of total art sales are useful in explaining the current total art sales.
In a similar manner to the previous regression, art total monthly sales are regressed on
adjacent closing prices of the S&P 500 index, as shown in Equation 4. The R2 and F-statistic are extremely small (R2 =0.08 and F-statistic= 1.5). This means that past prices of stocks are not good in explaining relationships with the current total sales in the art market. Due to this
reason of insignificance of obtained results, the second regression is dropped. My research
shows that there is not much of a direct relationship between art and financial markets. They
do show similar responses when faced with same economic shocks (Figure 6), but there is no
close relationship between them. Taking my investigation as an example, there is no clear,
direct or causal relationship between these two markets. Certainly, this may result from the
fact that I am comparing total sales in the art market to prices of stocks in the financial
market. The previous research has compared indices of prices of art and stocks and this may
R-squared 0.857581
Adjusted R-squared 0.849079 S.E. of regression 1.18E+08 Sum squared resid 9.35E+17 Log likelihood -1437.856
F-statistic 100.8611
Prob(F-statistic) 0.000000
Table 6 Reported results for the regression of Equation 8.
be why their results were different than mine. Moreover, my data selection might have been
imperfect. Sotheby’s sales might not be a good representative for the art market and S&P 500
might still be insufficient in capturing the whole financial market.
Next regression brings more illustrative results. It is art sales regressed on their lagged
values and lagged values of stock’s prices as in Equation 8.
Table 6 and Table 7 present results for the regression of art sales on both lagged values of art
sales and values of stocks’ prices.
In Table 6, Time Lags 1-4 (A) are the values of and Time Lags 1-4 (S) are the values of Equation 8. As observed in the first regression, the current total art sales are not significant in the equation for the total current art sales (p-value of Time Lag 1 (A)= 0.00).
Variable Coefficient Std. Error t-Statistic Prob.
C 49274442 93145565 0.529005 0.5987
Time Lag 1 (A) -0.163680 0.033959 -4.819867 0.0000 Time Lag 2 (A) 0.058190 0.023677 2.457670 0.0167 Time Lag 3 (A) 0.019441 0.001149 16.92036 0.0000 Time Lag 4 (A) -0.002333 0.000808 -2.888055 0.0053 Time Lag 1 (S) 90149.66 32788.31 2.749445 0.0078 Time Lag 2 (S) -15912.68 27215.32 -0.584696 0.5608 Time Lag 3 (S) -7305.323 2237.976 -3.264254 0.0018 Time Lag 4 (S) 160.9575 1114.389 0.144436 0.8856
Equation 8 are the values estimated on the basis of 12 lagged values of the Sotheby’s total sales and s are estimated values of the coefficients from the Equation 5 and where are the values estimated on the basis of 12 lagged values of the monthly stock prices of S&P500 and s are estimated values of the coefficients from the Equation 5, is therefore the predicted value of from Equation 5, here presented in form of a 3rd degree polynomial. is a constant and is the error term.
R-squared 0.888178 Adjusted R-squared 0.873978 S.E. of regression 1.08E+08 Sum squared resid 7.34E+17 Log likelihood -1429.149
F-statistic 62.54921
Prob(F-statistic) 0.000000
Table 7 presents very high R2 and F-statistic, which mean that lagged values of total art sales together with lagged values of prices of stock are good in explaining the current sales in the
art market. Stock prices, when combined in one regression with previous art total sales, add to
predictability of current sales in the art market, as the R2 value is bigger, when the third regression is compared to the first one.
Equation 9 presents the fourth and last regression executed, which is based on the
relationship between differences between consequent months’ values of variables.
Table 8 shows that first, second and third coefficient of the lagged value of the total art sales
month’s differences are insignificant in the equation.
Table 7 Descriptive statistics for the regression of art total sales on lagged art total sales and stock prices
(Equation 8).
Equation 9 are the values estimated on the basis of 12 lagged values of the Sotheby’s total sales month’s differences and s are estimated values of the coefficients from the Equation 6 and where are the values estimated on the basis of 12 lagged values of the monthly differences in stock prices of S&P500 and s are estimated values of the coefficients from the Eq. 6, is therefore the predicted value of from Eq. 5, here presented in form of a 3rd degree polynomial. is a constant and is the error term.
Variable Coefficient Std. Error t-Statistic Prob. Time Lag 1 (AD) 0.031388 0.142254 0.220651 0.8261 Time Lag 2 (AD) 0.235490 0.030370 7.754126 0.0000 Time Lag 3 (AD) 0.017481 0.002877 6.077274 0.0000 Time Lag 4 (AD) -0.007917 0.000958 -8.266977 0.0000 Time Lag 1 (SD) 25398.79 93431.68 0.271843 0.7866 Time Lag 2 (SD) -60077.65 40609.87 -1.479386 0.1441 Time Lag 3 (SD) -5426.841 4455.338 -1.218054 0.2278 Time Lag 4 (SD) 565.5519 1643.771 0.344058 0.7320
Table 9 brings a special result of an exceptionally high R2. This means that changes in art market sales are very well explained by changes in their past values of sales and changes in
stocks’ prices in following months.
4. Discussion and Conclusion
This research aimed to investigate the patterns of the art trade. The trends of the
demand in the art market were compared to those in the financial market. The investigation
R-squared 0.923354
Adjusted R-squared 0.913464 S.E. of regression 1.08E+08 Sum squared resid 7.18E+17 Log likelihood -1408.996
F-statistic 93.36393
Prob(F-statistic) 0.000000 Table 8 Reported results for the regression of Equation 9.
Table 9 Descriptive statistics for the regression of monthly difference of art total sales on lagged values of art sales monthly
focused on changes in the demand for art sold at Sotheby’s and on changes of the stock prices
determined by S&P 500 price index. Declines in stock prices were explained by declines in
the demand for them. Correspondingly, declines in the demand for art sales were related to
overall decline in prices in the art market. The idea to compare these two markets came from
the fact that stocks and art can be treated as financial assets. The differences in characteristics
between these two asset classes were described earlier. The different qualities of these two
investment opportunities might have resulted in alternate patterns of investment. What this
means, is that when investors observed worsening prospects for returns in the financial
market, they might have decided to transfer their funds to the market for art, where assets do
not exhibit systematic transfers with the assets in the financial market. Along this way of
reasoning, art could be treated as a hedge against downturns in the financial market. This is
why the hypotheses of Campbell (2008) were taken into account and attractiveness of art as a
way of portfolio diversification was examined. As art highly contrasts with stocks in its
asset’s properties, it is understandable to be treated as an alternative investment. Both
Campbell (2008) and Mei and Moses (2002) claim that art price index, when compared with
stock prices, is less correlated to assets in any other market. Such a characteristic would
support recommendations for constructing investor’s portfolio based on these two asset
classes. They would then balance investors’ portfolio returns by opposite reactions to same
economic shocks. On the other hand, Goetzmann (1993) and Channel (1995) implied that art
couldn’t be used as a hedge for financial stocks performance. Both of their researches found
that art and stock prices were highly correlated. Moreover, Granger tests performed by both of
them showed that art market never caused changes in the stock market, and the relationship
between them flowed from the financial to the art market, including certain lags. The delayed
behaviour in the art market was also explained by a lower degree of liquidity of art assets
Even though my investigation did not look at prices in both of the markets, but took
into account the realised demands of investors at Sotheby’s, it did showed similar findings to
the above researches. Just by looking at Figure 6 with combined charts of art total yearly sales
and average scaled yearly stock prices, it is visible that demands in both markets exhibit
similar movements. Parallel movements of demands in the art and financial markets are a
proof of a relationship that they both appear to have. Moreover, both of markets experience
similar consequences of the 2008 economic crisis. My paper focuses on the demand
characteristics of the two markets. I have therefore allowed myself to treat stock prices as a
representation for the financial investors’ demand. My findings show that financial market
reacts earlier to the economic shock. Average stock prices start to fall already in the end of
2007, while the total art sales at Sotheby’s respond to the crisis only after approximately one
year. As expected, art market demonstrates a delayed reaction to the economic shock. It does,
however, follow the trends of the financial market. What I conclude is that the demand in the
art market might be predicted by the behaviour of the demand in the financial market. When
there emerges a bear market in a trade in stocks, it is most probable that the situation will turn
in the same direction for the sellers in the art market. Even though the art and financial
markets do not show close bonds and neither there are any direct transfers between them, the
investing behaviour of buyers in these two different markets corresponds to each other. My
further findings present that current total art sales can be effectively predicted based on their
past values with stock prices’ lagged values. I propose a statement that the art market trends
are related to the financial market. I do not argue for a causal relationship, but I claim that
trends of the demand in the art market are in a retarded coincidence with the trends of the
demand in the financial market. I base my findings on the examination of the Global
Economic Crises of 2008. As financial investors drop their demand for decreasing in value
last economic crisis shows that bear markets emerge in the art market during major economic
recessions. Demands in the art market fell, as financial investors feel less wealthy. Just after
the rise of the anxiety in the financial market and the first symptoms of the 2008 crisis, the art
market experiences downturn too. I conclude that the most feasible reason for these results is
not necessarily causality or correlation, but a difference in degrees of liquidity. As paintings
are one of the least liquid assets, trade in them cannot respond to economics shocks as quickly
as that of financial stocks. Stocks being one of the most liquid assets immediately respond to
changes in the economy, whereas trade in art, which involves high transaction costs and much
time, is prevented from closely following the situation of the financial market. It is an
important finding and my research proves that art and financial markets are not each other’s
investment alternatives. Many would think that when a situation on the financial market
worsens, people turn to investing in dissimilar assets such as art, which are predicted to
respond to the economic situation in different way. It turns out that it is not the case, and
investing behaviour in both of these markets is relatively similar. The conclusion is that
crashes on the financial market will be followed by declines of the activity in the art market.
There exists a certain degree of predictability of the art market by performance of trading in
stocks.
On the other hand, the results of the regression of the art sales on past prices of stocks
showed that prices and thus the demand for stocks in the financial market couldn’t estimate
the sales in the art market. In spite of corresponding to financial market trends of the art
market, no direct casual relationship can be asserted. The results of this regression are
definitely not only the effect of imperfect variables selection. Some might claim that the
proposed lack of correlation between art and stock assets results from my decision to compare
studies, and the corresponding demand’s behaviour with delayed reactions is probably the
result of differences in liquidity characteristics.
Looking at my theoretical framework and my empirical results I come to the
conclusion that there is no direct relationship between these two markets. Even though the
demand trends correspond to each other, no causal relationship can be argued. The art market
follows the financial market, what is mostly caused by differences in the degrees of liquidity
between assets in these two markets. In my opinion, the results present a very interesting
conclusion that the market that is more liquid anticipates demand trends in less liquid
markets. In the case of art and financial markets, the reaction to the economic downturn is
first exhibited by the financial market, whose trends are closely followed by the market of art.
An important point to make is the fact that investors’ spending patterns in art and financial
markets are homologous. However, complexity of the art market needs to be kept in mind.
Especially the supply side and artworks’ valuations cannot be clearly explained. This is why
instead of prices, total values of sales were taken into account. Supply side is the most
problematic part of the art market. It might have been the case that it was not a decreased
demand of art investors but maybe rather lack of supply of desired artworks. Even though art
market’s trends are of such huge resemblance of those in the financial market, where
everything can be quantified and exchanged for money, art as an asset is still a category of its
own. In auction houses, the probability of purchasing an artwork only for financial gains is
very low. Art is still treated as a luxury investment, whose aesthetic value is most often more
important than the possibility of financial gains in the future. Nevertheless, many art funds are
emerging and their presence in the art market is the expression of its purely financial side.
For the future comparison of art and financial markets one should take into account the
limitations that my research has met. The improvements should regard the data selection.
should be researched. With huge progress in communication technology, data from all art
commercial institutions should be gathered. Sales volumes and values should be augmented
by price variables. The time period analysed could be extended and should always include the
most recent statistics. My research could be also improved by accounting for changing
consumer’s price index and differences in currency exchanges. The area of the art market that
needs the most of the researchers’ focus is its supply side. Influences of the supply of trade in
art should be clarified.
Notwithstanding, this work has some important implications for future investing.
Firstly, only the art funds should be treated as a reasonable way for creating a diversified
portfolio. Without expert knowledge, treating investment in art as diversification opportunity
might be quite mistaking. Moreover, during the periods of decreased demand for financial
stock, the demand for art decreases too. My conclusion is that art market does not pose an
alternative to the financial market. When investors withdraw their funds form stock
investments they will analogically cut their spending on art too. No casual relationship exists
between these two markets. What is true is the similarity of trends that they exhibit. As no
systematic transfers occur between art and financial markets, prices and demands for their
assets change independently of each other. The lesson for a future investor is to treat investing
in art market with precautions. Even thought one might think that art and financial markets
highly contrast with each other, my research finds out that in the end they do exhibit similar
patterns. This papers shows that decisions about portfolio diversification with art assets
should be made with precautions, as trends in the art market clearly exhibit same tendencies
as those in the financial market. This study with its focus on the latest time period brings out
the conclusions that demand’s behaviour in the art market declines with an economic
financial market. Art and financial assets should not be treated as substitutes but rather as
Bibliography
Campbell, R. (2008). Art as Financial Investment. Journal of Alternative Investments,
10, 64-81.
Campbell, r. (2005). The ‘Art’ of Portfolio Diversification. University of Maastricht
Thesis.
Candela, G., Scorcu A. E. (1997). A Price Index for Art Market Auctions. An
Application to the Italian Market of Modern and Contemporary Oil Paintings. Journal of
Cultural Economics, 21, 175-196.
Chanel, O. (1995). Is Art Market Predictable? European Economic Review, 39,
519-527.
Frey, B. S., Eichenberger R. (1995). On the Rate of Return in the Art Market: Survey
and Evaluation. European Economic Review, 39, 528-537.
Goetzman, W. N. (1993). Accounting for Taste: Art and the Financial Markets over Three
Centuries. American Economic Review, 83, 1370-1376.
Mei, J., Moses M. (2002). Art As An Investment And The Underperformance Of
Masterpieces. American Economic Review, 92(5), 1656-1668.
Mei, J., Moses M. (2005). Vested Interest and Biased Price Estimates: Evidence from an
Auction Market Journal of Finance, 60(5), 2409-2435.
Pesando, J. E. (1993). Art as Investment: The Market for Modern Prints. American
Economic Review, 83, 1075-1089.
Renneboog, L., Spaenjers, Ch. (2013). Buying Beauty: On Prices and Returns in the Art