The price of beauty : determining the price of paintings

Bachelor Thesis

10004084 Supervised by

M. J. G. Bun

Faculty of Economics and Business Universiteit van Amsterdam

2014

The Price of Beauty

Abstract

Many factors determine the price of a painting; most of them are irrational and numerically immeasurable. This paper focuses on the objective factors that can influence the price of an art work, in particular on the manipulative power of the art galleries. Art market is viewed from the perspective of a durable good monopolist. The annual growth of the price of paintings is found to be declining over years. “Death effect” is taken into account as well, and is once more proven to have an impact on the price of the art, in this case, paintings.

2 Table of Contents

1 Introduction 3

2 Literature Review 4

2.1 Empirical Literature Review 5

3 Sample Design 7

3.1 External Validity 8

4 Economic Theory 9

4.1 Justifying Art Market Inefficiency 9

4.2 Coase Conjecture 10

5 Data Analysis 12

5.1 Empirical findings 15

5.2 “Death Effect” 16

6 Sensitivity analysis 17

7 Conclusions and Suggestions for Further Research 19

8 Appendix 21

3 1 Introduction

A painting of two men playing cards sells for 259 million dollars; there are at least five countries with the nominal GDP lower than that (World Bank, 2012). What makes it possible to make such a comparison? Since the price of a good is determined within the market, price that high suggests that demand is either very high or very price-inelastic. However, it is a little bit different in the art market. The price of the painting is not set merely by the supply and demand but also influenced by a person – either an artist or his dealer. An artist can choose not to sell his work; he can choose not to sell it to some people. This discrimination will increase the price for the ones that are willing to purchase the art work. A good dealer is able to boost the price of a masterpiece using only subjective reasoning or no reasoning at all; when it is beauty that is for sale, who can argue about its value?

Furthermore, good dealers usually represent large auction houses, and their famous names attract more rich buyers and collectors than auctions held in smaller and less known galleries (McAndrew, 2013). Well known dealers are believed to have an eye for art and that also adds to the higher demand for the paintings sold at the world-famous auction houses, which in turn charge considerably higher price for the paintings. This does not mean that supply and demand determined in the market, as described in economic models, play no role in setting the price of an art work but rather that both can be manipulated. Indeed, the extent to which the prices of paintings are manipulated by the galleries would be simply illegal in most industries (Schrager, 2013). Almost all primary sales of the paintings are conducted by the art galleries (Schrager, 2013), making it easy for them to ignore the expectation that demand of the paintings is mostly determined by its aesthetic value, and create the demand that benefits galleries the most. One could say that galleries set not only the price of the painting but the taste for it too.

Supply side, on the other hand, has more of an economic rationale behind. Art is a durable good; hence, once supplied to the market, it will stay there, presumably, for ever – unless it is intentionally or accidentally destroyed. It seems then reasonable to assume, that if an artist produces less paintings during his lifetime, they will have a higher value after his death. One controversial example is William Paskell: during his lifetime he was already widely recognized as a very talented artist, however, in order to sell more paintings to provide for his large family, he became a very prolific artist, and even painted under different names to produce and sell more

4 paintings. The price of Paskell paintings is therefore low, however, it is rapidly appreciating as his works are being widely recognized (Woodward, 2012). This case confirms the excessive supply influence on a lower price of the paintings but the current increase in price of Paskell’s works has nothing to do with supply; since the artist is already deceased, supply of his paintings does not change. It is therefore difficult to predict the price of the painting and that implies that on both supply and demand sides there are other factors that influence pricing of paintings, other than mere aesthetic value and the quality of an art work.

The aim of this paper is to investigate what factors drive the price of the paintings up by examining the significance of their influence on the price of the painting. There are many possible drivers that are subjective and immeasurable, for instance, sentimental value, aesthetic image, particular taste etc; there can be no doubt that all subjective qualities have a great impact on the price since the primary purpose of art is to please an eye and sentiments. However, the ones that can be measured or examined are less subjective; these are the factors that are further studied in this paper: the year in which the painting was created, the year in which it was sold, the price paid, the name of the artist, the art movement and style, the city in which the art work was sold; some personal factors are also included in the research, defining the qualities that artists might have shared and that might have had influence on their genius.

2 Literature Review

Price fluctuations are quite unpredictable in the art market since they are not a subject to income or production costs of a firm; nor can an objective cost-benefit analysis be made (McAndrew, 2013). Between 1957 and 2007 art has appreciated in value by approximately 3,97% per year in real U. S. dollar terms (Renneboog and Spaenjers, 2011). Art works are seen as investment and studies are conducted to investigate its return and risk. Most of the papers and profound researches focus on art pricing from the perspective of an investor, arguing that art works do not underperform in the market but are a rather risky investment (Renneboog and Spaenjers, 2011).

The art market can also be analyzed from the point of the durable goods monopolist (UBC).This theory is based on the famous Coase conjecture about the patience game the monopolist (art galleries) and the consumers (the buyers of art) play in order to gain more surplus. Monopolist sets high price and expects to sell his good in the first period. Consumers,

5 who have high valuation of this good, are ready to pay that price. However, not all demand has been satisfied yet, and in order to sell more, monopolist has to lower the price in the following period. The more patient the consumers are, the more monopolist has to lower his price because it takes more periods for him to sell his good. Monopolist is competing with his future self in pricing his own good. Nevertheless, when art galleries engage in this kind of competition, they have manipulative measures to win. Dener (2007) goes into more detail about durable goods monopoly and Coase theorem in his working paper on quality uncertainty and time inconsistency.

2.1 Empirical Literature Review

There exists a large empirical literature, aiming at determining the variation of the price of paintings. Stein (1977) and Baumol (1986) took auctioned paintings as a random sample of the stock of art (only by deceased artists) and based their research on yearly average sale price. Pommerehne and Frey (1989) used a geometric mean to calculate the average rate of return per period on an investment that was compounded over multiple period on works that were sold two or more times within the selected period. The average of the annual growth determined by the previous researchers was above 10 per cent. Return on paintings was found to be fluctuating a great deal because of the varying prices. McAndrew (2013) found that the fame of an artist is an influential factor. Ekelund, Ressler and Watson (2006) argue that after the death of an artist, the supply of his paintings becomes finite because the “output” of an artist – a durable goods monopolist – ends. Value of his paintings therefore increases – this is known as the “death effect”. Matheson and Baade (2003) discuss the “death effect” from the perspective of the durable goods monopolist and add additional reasoning for it, called “nostalgia effect”. The latter is mostly caused by the media and overall attention to the deceased artist right after his death, increasing interest in his work as a result.

“People are spending millions on works by artists who have questionable long-term value”, says Ivor Braka, London dealer, buying and selling high-value paintings since 1978. The future value of a painting cannot always be predicted with certainty; yet a lot of people purchase art works as investment. The main reason indicated by the Renneboog and Spaenjers is that investors seek to diversify their portfolios by adding collectables. In their paper they argue that

6 high value art more often results in higher returns, however, art market is very volatile and therefore its performance is comparable to the corporate bonds market – at a much higher risk.

Art market has its own cycle of booms and recessions and is not irresponsive to financial market ups and downs (Chazen, 2012). However, in his book “Talking Prices” Velthuis points out that art market crash in 2008, which somewhat coincided with the financial market crash, could have not entirely been due to the latter. Art market, he argues, has already been so overheated that the trust in the contemporary art investments dropped approximately 40 per cent even before the financial crisis begun. Since the 2008, the share of the paintings by well-established artists, such as Bacon, Warhol, Kooning or Basquiat, in the auctions increased from 46 per cent in 2008 to 63 per cent in 2011 (Petterson, 2011). That can conclude that paintings by well known artists are a safer investment, even witnessed during the recession. Expectation therefore would be that the name of an artist and ability to prove the authenticity of the painting will increase its price.

Renneboog and Spaenjers (2011) conducted a significant research on the returns on paintings using a sample of more than a million observations. They found that art returns (and, therefore, prices) among other things depend strongly on consumer sentiments; the R-squared of their hedonic regression came up to 0.52 which is quite high for the fluctuating art market. Renneboog and Spaenjers (2011), however, criticize most of the previous researches, since neither of those methods takes the variation of quality into account – there are at least three problems with the methods used. Firstly, paintings are not traded frequently – that gives a quite small number of observations for a chosen period, and every new observation will influence the estimators (Mei and Moses, 2002). A larger sample is beneficial because both more observations and more variation lead to increased precision of estimates. Secondly, it is impossible to avoid selection issues. For example, samples of the repeated sales are quite location-biased because the first transaction might occur at any auction house but if the artist becomes famous or the painting is very popular, such giants as Christie’s or Sotheby’s will almost certainly hold the second auction. Finally, even the works that trade more than once might not be representative of the all population of existing paintings.

7 To avoid the first problem, this research was not based on the number of the repeated sales only but on all works by the selected artists that have ever been sold, including repeated sales. That yields a larger sample, and therefore may lead to increased precision. In order to tackle the second issue, location was included in the regression as a dummy variable, with a value of one for the large and famous auction houses, and zero otherwise. Every sale is recorded with the sale location, indifferently of whether it was the first sale or not, location bias is thus eliminated. The complete solution to the last problem is to take the whole population as a sample, which is time consuming and quite difficult to achieve, and therefore is not used in this research.

3 Sample Design

The sample consists of 7446 observations of sales of the oil-paintings and works on paper by selected artists; this choice of medium was made due to the fact that these artworks account for approximately 85 per cent of the total turnover in the art market (Renneboog and Spaenjers, 2011). The purpose of this study is to answer the question what determines the price of a painting, therefore, the sample was selected using the following reasoning. Ten artists were selected and all their paintings that have been sold, including repeated sales, were taken in the sample and described with regard to the factors that are the parameters of interest in this research. The criterion used to choose those particular artists was the total value of the paintings ever sold, including repeated sales, in real U. S. dollar terms, adjusted for inflation, i. e., of the value of the year 2014. The rationale behind this choice is that whatever those ten artists (or the sales dealers) did to achieve people paying the price that high, had to contribute to the increase of price. Hence, all the properties that paintings that are in the sample share, have to be tested for significance and if they are statistically significant, model can be applied to any painting that is not in the sample. To put it differently, if a painting that is not in the sample has similar characteristics that are proven to be significant in determining the price of a masterpiece, it is expected to be similarly expensive.

The sample is large enough to draw reasonable conclusions from the results of the regressions; also, since all paintings of the included artists, not only the famous or most expensive ones were taken into the sample, data presents an opportunity to examine the relationship between the price and its influential leverages in the presence of subjective and

8 innumerable factors. That is because all chosen artists have a distinct style and technique, a possible bias that might occur due to the preference of a certain manner of drawing is eliminated since the buyers that have a particular liking still chose to pay an excessive price for some paintings of one artist but not for the other works by the same master. That gives a parallel insight that the fame and recognition of an artist are established alongside with the pricing of his works, and not before; which can also be supported by an argument that for an artwork to become famous it has to be seen by many people, thus the more time the painting is exhibited, the higher price it should attract (McAndrew, 2013). That is the reason why the repeated sales were included in the sample as well.

3.1 External Validity

The pitfalls of the way the sample was designed are not to be overlooked either. The sample size is approximately 150 times smaller than a predictable population size that includes the paintings of all artists ever recorded to be sold at the auctions. This model is expected to be able to explain the pricing of the paintings that share similar characteristics with the paintings in the sample. For instance, if the most expensive paintings share the same properties, such as the particular art movement, brand new ideas, or are made popular due to some special features of the artist, and those properties are statistically significant, the paintings that are not in the sample but have the abovementioned characteristics, could be expected to bear the higher price than the ones that do not share those properties. However, the shade of doubt befalls the actual significance of the properties – if they are proven to be influential within the sample, it does not mean that the same significance level will hold externally. Moreover, sample is not entirely random, since the ten artists were picked for a reason that is presumed to be valid for the reasons mentioned earlier in this section. If a new style of art emerges, and the paintings of that style sell well, sample needs to be adjusted in order to make the model more accurate; in other words, model based on this sample is not able to adapt to completely new information on paintings and their sales. The model is therefore valid under the condition that the circulating paintings and the new ones that come into the auctions do not change much. Since art market is not only unpredictable in pricing but even more so in the possible “new products”, to increase the external validity of this model, sample could be more random, its size enlarged, as well as more influential factors should be taken into account.

9 4 Economic Theory

Schrager (2013) discusses high-end art market inefficiency and concludes that it is due to the enormous extent of price manipulation by art galleries. A significant part of her paper is dedicated to the manipulation of prices, why, how and who makes it possible. She argues that increasing the transparency of the galleries, coming back to the patron-model or holding more primary sales at the auctions could make art market more efficient and leave less space for the manipulation. Hence, the price of paintings would most likely decrease. McAndrew (2013) focuses on art market research and analysis; in her study of pricing Pablo Picasso paintings she attempts to explain the changes in prices of the famous artist’s paintings by employing linear regression. Although linear form is acceptable in performing diagnostic tests, McAndrew came to an obvious conclusion that multi-variable model is needed to explain more of the variation in the model. This study therefore is not taken into account from the econometric point of view but for the profound background research about the price influencing factors.

4.1 Justifying Art Market Inefficiency

The control of the art market is essential for the galleries to make profits; hence, it is kept tight. Potential buyers, represented artists and even their works for the exhibitions are carefully selected. Before the painting is shown, it has already been offered to preferred clients – museums or known private collectors (Schrager, 2013). Art galleries hold a grip on the secondary market by manipulating the auctions and private owners who sell the paintings in a quite simple way. If a private owner sells a painting to a gallery, it is made sure that he or she is cut off and can make no further purchases of the paintings that are within the reach of the gallery. Rationale behind this is that every sale reduces the price, which stems from the economic theory – an occurred sale can be seen as an increase in supply by one painting, and since obviously the owner is no longer willing to hold the painting, a decrease in demand by one painting. For the exact same reason galleries send their people to the auctions to bid the higher price, and thus increase the perceived value of the painting. It is not uncommon for the gallery owner himself to bid the higher price for the artist whom his gallery represents (Schrager, 2013). That makes perfectly clear why galleries choose not to sell to those who might sell the art work in the secondary market; dealers have to know the buyers. Reasoning behind this is again quite simple – keeping track of the painting

10 makes sure that it is available for future exhibitions and the value is not affected by the lower pricing in the secondary market (Schrager, 2013).

Art market inefficiency is caused by the galleries offering discounts to the prestigious museums and big collectors, and the exclusive opportunities to purchase art for the collectors who are able to influence public taste for the paintings. Brand building is also done by the galleries: fostering new prospective artists by visiting their studios, introducing them to the art collectors, and directing their further career (Schrager, 2013). However, inefficient market is not the only cause of the high prices of paintings. Well-known art dealer Marla Goldwasser indicates that aesthetic value is the most important criterion when pricing art. It does not have to be just “beauty”, although it definitely adds value if the painting is pretty to look at. Aesthetic value is also perceived to be high if the painting is provocative or very emotionally moving (Schrager, 2013). However, since it is not uncommon that the taste of the big collectors is set by the art galleries, aesthetic valuation can be manipulated as well. It should indeed be worrying that galleries that have the biggest financial interest are able to control the art market. Price manipulation in any other market would cause distortions, shortages and inefficiency; yet the “inefficient” art market generates tens of billions of dollars of revenue every year (Schrager, 2013).

4.2 Coase Conjecture

Inefficient non-public goods market is where either its participants lack information or experience major principal-agent problems, or where time-inconsistent preferences exist. The last characteristic is indeed common in the monopoly with durable goods. Art works are durable unless destroyed and not substitutable, and the major galleries have a tight grip over the market, therefore, it can be viewed from the perspective of a durable goods monopolist (UBC, 2010).

The gallery sets the price of a painting (directly or in the manner discussed earlier), and usually this price is quite high when talking about the artists that are represented by the large and famous galleries. There are some collectors and other people willing to pay such a high price for a piece of art because their personal valuation of the painting is very high. It might be that they find the painting particularly attractive, appealing to their memories or sentiments or simply have enough wealth to become famous for paying an incredible price for an art work. There are other

11 people, however, who do not value the painting as much or cannot afford to pay such a high price for it. The monopolist – the gallery – has to wait until someone buys the painting for the primary price or decrease the price and hope that there will be a buyer willing to pay that price. If nobody is willing to buy the painting yet, the price has to be lowered even more. Buyers know that the price will decrease if they wait but monopolist knows that buyers are ready to wait for the price to fall. This is the game of patience that is linked to the Coase conjecture, first published in 1972.

It is assumed in the Coase theorem that individual buyers have different valuations of the paintings and that galleries do not know those valuations. The galleries could sell the paintings for a low price but they want to maximize their profits and expect that there are some buyers who are willing to pay higher price. Assuming different periods of pricing, galleries will set the price really high for some paintings in the first period. Buyers who are ready to pay price that high will purchase paintings in the first period; however, if there are some paintings left, the galleries will have to lower the price and try to sell them in the second period for the buyers who have lower valuation of those paintings or simply cannot afford to pay the first period’s price. If not all paintings are sold in the second period, the price will be lowered again and galleries will expect to sell the art works in the third period; this continues until the price equals marginal cost and cannot be further decreased because that means loss for the seller. Galleries are competing with their future selves, deciding whether to lower the price or not. For all art buyers it is worthy to be as patient as they can, waiting for the price to decrease, running a risk, however, that somebody else will purchase the painting that they want to acquire: a problem of time inconsistent preferences.

12 In the first period at price p1 demand is satisfied for the part 01 on the demand curve. In the second period, when the price is lower (p2), demand is satisfied until point 2 on the demand curve; galleries can decrease the price until it reaches marginal costs, then they will experience loss.

However, since art galleries are competing against themselves, it is easy for them to win this battle. Manipulation techniques and market control guarantees that the price set by the gallery in the period one is the price for which the painting is sold (Dener, 2007). This makes art market different in a way from the other markets where durable goods monopolists engage in production; it is easier to create excessive demand through the aforementioned market manipulations for paintings, than for other durable goods, for instance, refrigerators. Once more one runs into the same subjective concept “aesthetic value” which is so strong and influential in the art market.

5 Data analysis

Previous studies and the preliminary research for this paper show that art galleries have a great impact on the art market. As described in former sections, the power of art galleries covers both supply and demand sides of the collectibles market. While conducting the primary theoretical research and data collection for this paper, it stood out that high prices of some paintings are often assigned to certain auction houses. This is also justified by Schrager (2013), as discussed in

13 section 4.1. Therefore a variable representing the auction house is carefully examined among the others. Most regressors were chosen based on the research by McAndrew (2013), discussed in section 4: year of creation, year of sale, style, a contribution by an artist and a location of sale. In this paper the location of sale is a variable of interest, while style (different periods of Picasso creation) was a variable of interest in McAndrew’s research.

The price of the painting is selected as the dependent variable, while year of creation, year of sale, style, sale location, as well as dummy variables indicating whether an artist had a mental disorder, was extravagant and created something entirely new or contributed to the art movement significantly, and finally, whether he was still alive when his works were for sale, were included as independent variables. Variable “style” was also made a dummy because more than half of the ten chosen artists are either impressionists or represent art movement that is a product of impressionism; the rest of the list falls into the modern art category. That implies that impressionistic works are more popular that others, moreover, they dominate the list of the most expensive paintings ever sold as well (The Art Wolf). Year of creation vary from 1855 to 2012, and the sales are analyzed from 1957 to 2014; latter is the period that this paper focuses on, defined separately as the year of sale.

Places of sale vary a great deal city-wise, and are of a particular interest, therefore the paintings that were sold at Christie’s or Sotheby’s auction houses are distinguished from the ones that were sold at different art auction houses; sale location variable, hence, is also a dummy. Reasoning behind this is that two aforementioned art auction houses are currently the world’s largest and the most luxurious art businesses and previous researches have shown how much an art gallery can influence the price of a painting. Furthermore, it is interesting to investigate the power of the price manipulation that galleries have in the art market. Based on the theory described in sections 2 and 4, a positive coefficient for sale location is expected.

Variables defining personal characteristics were taken from the reliable biographies of the artists. A dummy indicating mental disorder (“mental”) is 1, if an artist had one, and zero otherwise. Most of the artists with a known disorder were autistic. There are previous researches on the relationship between autism and artistic abilities. Buck, Kardeman and Goldstein (1985) argue that autistic artists are more sensitive to the surrounding world and put a greater effort to communicate through their works. Paintings therefore become more accurate in the sense of

14 alluring to people (Hermelin and O’Connor, 2006) and are more attractive to buy. The dummy “mental” has a positive coefficient of 0.98, indicating that autism is positively affecting the price of a painting, and a weak positive correlation with price (Table 2, appendix). Variables “mental” and “extravagant” have a positive correlation of 0.54. It could be expected that a sensitive and expressive person might choose to present himself in a rather unrestrained manner. Variable “new”, indicating whether an artist has significantly contribute to the art movement or created something new, positively correlates with both “mental” (0.33) and “extravagant” (0.54), which is not surprising since those two correlate with each other. The interesting insight here might be in accordance to the one of Buck et al (1985) – that autistic artists are more aware of their surroundings and may notice things that others would not. That is, it might be easier for them to create something new or exceptional.

In order to analyze data properly, some adjustments needed to be made. Since there are some significant price spikes, the estimation method will be less sensitive to those outliers. Therefore, a natural logarithm of the price is used in order to reduce the relative differences between prices of paintings in the model. Generally knowing that art market prices are without doubt driven up by new and exciting creations, the growth rate of the price of the paintings was suspected to be declining over time; or at least up until the next innovative revolution in art. Hence, it was checked for by sorting the year of sale and taking the percentage differences of the price of each year, and then estimating the intercept and the slope of the simple linear regression, taking the percentage differences of the annual price as dependent and the year of sale as independent variables. The result was equation (1), which shows that with every year the increase in price of paintings is indeed declining:

ŷ = 2.912 – 0.002x (1)

(0.29)

ŷ – percentage difference in the increase of price of a painting x – year of sale

Evidentially, the relation between the price and the year of the creation of the painting is not entirely linear. Therefore, one more independent variable, equal to the square of the year of sale, was included in the analysis to cover this non-linear relationship.

15 5.1 Empirical Findings

Regressing the natural logarithm of the sale price of paintings on the independent variables has shown that all the coefficients in the regression are statistically significant, both jointly and individually (Table 1, appendix). Most variables in the regression were adjusted in order to avoid the differences of their scale, by equalizing their scale of measurement: taking either a dummy variable or the natural logarithm. Only the year of sale and the year of creation were left untouched; the natural logarithm of price was therefore meant to reduce the influence of price outliers. Most influential variables, both correlation-wise and having the largest coefficients and highest F-statistics on them, are year of sale and sale location, with a positive impact on sale price, and dummy variable indicating whether there was a significant contribution to the art movement, which has a negative coefficient in the regression. It is straightforward with the year and location of the sale, meaning that the later the painting was sold (including repeated sales and adjusting for inflation), the higher price was paid for it, and that high price was influenced by the art dealers of a certain auction house (once more confirming the manipulative power of galleries). The correlation between the year of sale and the price is approximately -0.29, while the correlation between the location of sale and price is approximately 0.32 (Table 2, appendix). This indicates that the paintings that sell at large and well known art galleries, such as Sotheby’s and Christie’s are indeed likely to bear higher price. On the other hand, it is quite curious that the variable indicating contribution to the art movement has a negative coefficient, and correlates with price by -0.04, showing a very weak negative relation. This means that if an artist had contributed significantly or created something entirely new, the price of his paintings is slightly lower than if he had not. This might be explained by the higher variance since one can never be sure how to evaluate something completely new and can only predict if it is going to be a success. However, if the exceptional creation turns out to be a best seller, it should be more expensive than other “regular” works; the negative relation might only be induced by the variance due to the uncertainty whether the initiative is going to be appreciated by the general public.

Sample being quite large and all independent variable proving to have some effect on the price of paintings, the R-squared of the regression was 0.2734, meaning that 27.34 per cent of the total variance can be explained by this model. Although R-squared is not high, for

16 such a volatile market it shows that the model explains variance quite well. R-squared might be increased and the reasons could be twofold: either the log-quadratic model is too trivial to explain the determination of the price of the paintings, or the immeasurable variables, such as personal taste and particular liking, play more important role than the ones that can be measured in setting the price for a painting. Latter argument seems to be quite valid, knowing that one’s taste and perception of beauty cannot be measured. Adding a rational measure of taste would, without doubt, significantly increase R-squared.

These subjective evaluations together with some of the factors mentioned in section 4, for instance, the pricing behavior of the art galleries, could be also classified as omitted variables. Unfortunately, that would increase the probability of the omitted variable bias in the model. Subsequently that would cause an over- or underestimation of the included variables and inconsistent estimators.

5.2 “Death Effect”

An interesting finding from the previous studies, the significance of the “death effect”, is also present in this paper. A working paper by Ekelund et al (2000) studies the “supply-included” demand effect – the fact that the death of an artist is also an end to his “output”, from the perspective of a monopolist that produces durable goods. It is referred to as “supply-induced”, meaning that the demand for the paintings is boosted by the limited supply after the death of an artist. Moreover, researchers state that this is rather demand than supply phenomenon, since the reduced supply stimulates demand. The research was conducted for the period of 1977 and 1996 (which falls into the year interval that this paper focuses on), and a clustered increase in the price of the art works just after the death of an artist was observed. The base for their hypothesis was the famous Coase theory on the durable goods monopolist; they argued that the increase in the demand, and therefore price, for a painting is due to the anticipation of the death of an artist because that assures that the value of a painting will not be decreased by the increasing supply of the paintings of the artist. In this paper their argument can be confirmed by the finding that the dummy variable “dead” that indicates whether an artist has already been deceased at the time of selling his works indeed positively correlates with the price of sale by 0.29, implying a 29 per cent increase in sale price. Furthermore, in the regression variable “dead” is almost as influential as the location of the sale.

17 Matheson and Baade (2003) confirm the significance of the “death effect” in their study on collectible prices and add “nostalgia effect”. The fact, also encountered by Ekelund et al, that the price of collectibles increase immediately after the death of an artist but eventually falls over time cannot be explained by the durable goods monopolist theory. Since the supply of the paintings is fixed after the death of an artist, the increase in price cannot be explained from the supply side. It can be justified by the “nostalgia effect”, which was introduced for that purpose. Matheson and Baade (2003) compare art market with sports collectibles of the deceased athletes and find that items left by famous characters indeed confirm the “nostalgia effect” by selling for significantly higher price than during the lifetime of the owner (sports) or creator (art). The more well known the artist, the more media attention he will get after passing away and that will attract more interest in his works. Interest translated into demand and increased demand raises the price of the paintings by the deceased artist. One could say that “nostalgia effect” is a part of the “death effect”, fortifying the latter depending on the fame of the artist.

6 Sensitivity Analysis

R-squared being 0.27, while all variables are proven to be statistically significant with no perfect multicollinearity, indicates that a model can be improved. Firstly, one-at-a-time (OAT) approach was employed, and regression was run excluding one independent variable at a time to see the following changes in the model. This assures that the model is checked carefully and not a single variable has missed the check. After excluding variable indicating the year of creation of the painting, dummy variable indicating the style became insignificant, and coefficient of the dummy variable “new”, indicating the significant contribution to the art movement, became positive. Year of creation and style are correlated by -0.52, it is obvious, that each movement of art has its own time frame; therefore this finding is not surprising. However, the relation between the year of creation and style of a painting is not strong enough to eliminate the style from the initial regression; diagnostically regressing style on the year of creation, i. e., performing a simple linear regression, has shown that year alone explain less than one third of the style variance. Interestingly, after including the variable “dead”, even when the year of creation is excluded from regression, style does not become insignificant. That again proves the significance of the “death effect” and shows that it can compensate for the missing boosters of the price.

18 Excluding dummy variable indicating whether an artist was extravagant decreases the significance of the variable indicating a presence of a mental disorder but decreases R-squared by just 0.0018. Dummies “extravagant” and “mental” have quite strong positive correlation as has already been mentioned before; nevertheless, there is no multicollinearity deducted. Once more, when a dummy “dead” is in the regression, “mental” does not become insignificant after eliminating “extravagant”. Excluding the location of the sale alone decreases R-squared by approximately 7 per cent; art dealers really seem to be influential on the price of the paintings. Without the galleries controlling the art market, the price of the paintings would be without doubt lower.

When regressing artists one by one, i. e., testing the impact of the variables on price per artist, the problem of multicollinearity arises. “Style”, “mental”, “extravagant” and “new” are omitted due to collinearity. R-squared varies greatly for each artist, and some variables become insignificant, as seen in the table on the next page. Individual regressions per artist do not take “death effect” into consideration, since variable “dead” will be as well omitted due to multicollinearity.

Artist R-squared Less significant

Andy Warhol 0.0852 sale year

Pablo Picasso 0.1206 sale year

Paul Cezanne 0.1660 year of creation

Vincent Van Gogh 0.4791 -

Jackson Pollock 0.1298 year of creation, sale year, sale location

Francis Bacon 0.2219 sale year

Claude Monet 0.1066 year of creation

Willem De Kooning 0.1225 sale year

Gustav Klimt 0.1773 -

19 Model is capable of explaining the variation of price of Van Gogh paintings relatively well compared to other artists in the sample, and seems to be almost unable to predict the discrepancies of the price of Warhol’s works. This observation may be attributed to the fact that although both artists were committed to one (but obviously not the same) art movement, expressionism, represented by Van Gogh, is more predictable in terms of medium, style, manner of drawing and theme of a painting than modern art, represented by Warhol. Taking the “death effect” into account, the parameters for Van Gogh might have been steadier over time of the sales of his works, since he was not alive when Christie’s and Sotheby’s began auctioning his works; and this contributes to a smaller variance of the price of Van Gogh’s paintings.

When correlating the prices of the paintings among artists in the sample, no significant relationship has been found. Interesting finding could be that prices of the paintings by artists that have introduced something new for their time, like Warhol, Picasso, Cezanne and Van Gogh have mostly negative weak correlation with the prices of all other artists (Table 3). That indicates that although prices in the art market move together, price increase for paintings by particular artists or their properties might decrease the price for other paintings. On the other hand, it is quite possible that this is merely one more prove of the manipulative galleries power over the price of paintings. Setting the taste of big and influential collectors which is a proven practice by the art galleries (Schrager, 2013) is likely to induce a negative correlation between the prices of paintings by different artists. If it is beneficial for a gallery to sell the works of Picasso, and large collectors demand his paintings, it is quite obvious that the paintings of similar style will increase in price while the opposite will happen to, for instance, modern art.

7 Conclusions and Suggestions for Further Research

Art market is quite unpredictable in almost all aspects it contains: the new creative ideas, the taste for the collectives, the quantity of art works, and naturally, prices. Under a tight grip of the art galleries, controlling supply and demand, and even setting the taste of collectors, this market satisfies the criteria of an inefficient market. Nevertheless, it is growing and earning tens of billions of dollars of revenue every year (Schrager, 2013).

The price of the art works is fluctuating, responding to both wimps of galleries and financial market movements (Chazen, 2012). The price of paintings is hardly predictable;

20 however, there are some elements that seem to have some impact over it. Despite the immeasurable perception of beauty, undoubtful power of galleries and extraordinary autistic talent, accompanied by the clear sympathy for impressionism and its sequels are the main factors that influence the price of the paintings. “Death effect” is once more proven to influence the price of the paintings. The increase in price of an art work after the death of an artist is explained by the end of artist’s supply and therefore the increased value of his works, and by the increase in demand of his paintings due to the exclusivity that the paintings now posses. Enhanced by the “nostalgia effect”, “death effect” strikes immediately after the death of an artist, and fades away over time (Matheson and Baade, 2003).

The model is able to explain approximately 27 per cent of the total variance of the prices of paintings, all independent variables are proven to be statistically significant and the sample used is quite large. However, the external validity of the model implies some suggestions for the future research in this field. First of all, sample should be more randomized and enlarged, which is always a secure way to increase the reliability of the model. In this model ten artists that have the highest total value of their paintings ever sold have been chosen. Hence, they have something in common, as mentioned in section 3. As an alternative, a totally randomized drawing of the artists from the whole art data base would be an option; however, that would have to be a very large number of artists to represent the whole population. The prices of paintings could also be studied from the point of each aspect separately, namely, when a certain art movement, time period, artist himself etc is chosen. Secondly, the correlation among the prices of individual artists could be conducted by taking the price indices of each artist and examining the relationship among them. In order to do that, a price index for each artist should be derived and calculated. These indices could also be compared to the art market index, once it is calculated. That would clarify the movements of the prices of paintings within the market and the degree of control by the art galleries. Finally, the rational measures of taste, appeal to personal feelings and sentiments and evaluation of aesthetic value included in the regression would carry a significant weight in determining the price of paintings. Measuring those subjective valuations is the key to a more transparent art market which would lead to the overall decrease in prices of paintings and, as a matter of fact, other collectibles.

21 8 Appendix Table 1 Regression

β1 Std Err.

Creation Year -0,013 0,002

Style -0,874 0,077 R-squared 0,2734

Sale Year 5,331 0,601 Root MSE 2,0366

Sale Year 2 -0,001 0 N 7446 Sale Location 1,666 0,062 Mental 0,978 0,081 Extravagant -0,542 0,084 New -1,297 0,105 Dead 1,504 0,158 Table 2 Correlation Sale Price Creation

Year

Style Sale Year Sale Year 2

Sale Location

Mental Extravagant New Dead

Sale Price 1 Creation Year -0,289 1 Style 0,093 -0,521 1 Sale Year -0,120 0,595 -0,360 1 Sale Year 2 -0,121 0,596 -0,361 1 1 Sale Location 0,318 -0,108 0,120 -0,123 -0,123 1 Mental 0,033 -0,093 0,350 -0,131 -0,131 0,305 1 Extravagant -0,214 0,021 0,014 0,038 0,039 0,079 0,542 1 New -0,041 -0,512 0,358 -0,226 -0,226 0,041 0,331 0,535 1 Dead 0,290 -0,903 0,471 -0,631 -0,632 0,060 -0,073 -0,197 0,486 1

Table 3 Correlation among Artists (Price)

Warhol Picasso Cezanne Van Gogh Pollock Bacon Monet Kooning Klimt Richter

Warhol 1 Picasso 0,021 1 Cezanne -0,011 -0,011 1 Van Gogh 0,021 -0,030 -0,016 1 Pollock -0,019 -0,018 0,050 -0,018 1 Bacon -0,018 0,070 -0,018 -0,037 -0,050 1 Monet -0,029 0,154 -0,034 -0,051 0,012 0,003 1 Kooning -0,006 -0,001 0,025 -0,032 -0,028 0,129 -0,047 1 Klimt -0,009 -0,007 -0,010 -0,019 0,007 -0,016 0,030 0,009 1 Richter -0,017 0,002 -0,026 0,004 -0,045 0,062 0,044 0,241 -0,007 1

22 7 Bibliography

McAndrew, C. (2013, May 1). A Study of Picasso. Art Prices. Lecture conducted from Trinity College, Dublin.

Buck, L. A., Kardeman, E., & Goldstein, F. (1985). Artistic Talent In "Autistic" Adolescents And Young Adults. Empirical Studies of the Arts, 3(1), 1-1.

Caines, M. (2014, March 20). International art market 2013: new report examines the facts and figures.The Guardian. Retrieved July 3, 2014, from http://www.theguardian.com/culture- professionals-network/culture-professionals-blog/2014/mar/19/international-art-market-2013-facts-figures

Chazen, D. (2012, January 19). Cornell University. The Art Market Crash and the Global

Economy. Retrieved July 17, 2014, from

https://confluence.cornell.edu/display/tam2011/Day+18+-+Daniel+Chazen

Dener, E. (2011). Quality uncertainty and time inconsistency in a durable good market. Journal

of Economics, 104(1), 1-24.

Ekelund, R. B., Ressler, R., & Watson, J. (2006). The "Death Effect" In Art Prices: A Demand Side Exploration . Journal of Cultural Economics, 24(4), 283-300. Retrieved July 3, 2014, from http://econpapers.repec.org/RePEc:kap:jculte:v:24:y:2000:i:4:p:283-300 Fernandez, G. (2008, May 1). Most Expensive Paintings Ever Sold. Most Expensive Paintings

Ever Sold. Retrieved May 29, 2014, from http://www.theartwolf.com/10_expensive.htm

Hermelin, B., & O'Connor, N. (1990). Art and Accuracy: The Drawing Ability of Idiot-Savants. Journal of Child Psychology and Psychiatry, 31(2), 217-228.

Matheson, V. A., & Baade, R. A. (2004). 'Death effect' on collectible prices. Applied

Economics, 36(11), 1151-1155.

Mei, J., & Moses, M. (2002). Art as investment and the underperformance of masterpieces. United States: American Economic Review.

Petterson, A. (2011, September 12). Is art still a safe bet for investors?. The Art Newspaper. Retrieved June 24, 2014, from http://www.theartnewspaper.com/articles/Is-art-still-a-safe-bet-for-investors/24508Renneboog, L., & Spaenjers, C. (2013). Buying Beauty: On Prices and Returns in the Art Market. Management Science,59(1), 36-53.

Renneboog, L., & Spaenjers, C. (2011).Buying beauty: on prices and returns in the art market. Tilburg: Tilburg University.

Schrager, A. (2013, July 11). High-end art is one of the most manipulated markets in the world. Quartz. Retrieved July 17, 2014, from http://qz.com/103091/high-end-art-is-one-of-the-most-manipulated-markets-in-the-world/

UBC. (n.d.). Durable Good Monopoly. University of British Colombia. Retrieved July 10, 2014, from http://faculty.arts.ubc.ca/pnorman/durablegoods.pdf

Woodward, L. (2012, August 1). Law of Supply and Demand for Artists.woodwardsimons.com. Retrieved June 24, 2014, from http://woodwardsimons.com/blog/47585/law-of-supply-and-demand-for-artists

Velthuis, O. (2005). Talking Prices: symbolic meanings of prices on the market for