THREE ESSAYS IN APPLIED MICROECONOMICS: ART PRICING, ART MUSEUMS AND ART AUCTIONS

By ARZU AYSIN TEKINDOR

A dissertation submitted in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY

WASHINGTON STATE UNIVERSITY School of Economic Sciences MAY 2013

To the Faculty of Washington State University: The members of the Committee appointed to examine the dissertation of ARZU AYSIN TEKINDOR find it satisfactory and recommend that it be accepted.

___________________________________ Vicki McCracken, Ph.D., Chair

___________________________________ Jill McCluskey, Ph.D.

___________________________________ Ana Espinola-Arredondo, Ph.D.

ii

Acknowledgements I would like to express the deepest appreciation to my committee chair, Dr. Vicki McCracken, for her support throughout my PhD program with her patience and knowledge, leaving me all freedoms for my research topic, encouraging me to follow my passion for art and believing in me when I questioned myself. Without her suggestions and valuable comments this dissertation would not have been possible. I would like to thank my committee members, Dr. Jill McCluskey, Dr. Ana EspinolaArredondo, for their constructive feedback, precious time and ideas for improvement of my study. Their interest and support was a major motivation. Many thanks go to my classmates and colleagues at the School of Economic Sciences for their kindness, friendship and support. I owe them deep gratitude for their help during my PhD program and feedbacks during my presentations. I would like to thank all my friends in Pullman and in Turkey for making me feel home away from home. Through my journey, I could not have done it without my biggest role models in my life, my mom and my dad. I express appreciation to my parents, Seher Tekindor and Ismail Tekindor, the first to believe in me and to encourage me to pursue my dreams. I appreciate their patience and understanding. I could not have gone so far in my life without their support. Many thanks go to my brother Gökşin and my cousin Didem for their confidence in me and help in my education. I also would like to thank my loved ones, who have supported me and always with me throughout my bad and good times. Things were so much easier and more fun when you were in my life. You will forever be appreciated. Finally, I am indebted to my art advisor, an outstanding artist, Hızır Teppeev. Without him, I would not have got the inspiration for this dissertation.

iii

THREE ESSAYS IN APPLIED MICROECONOMICS: ART PRICING, ART MUSEUMS AND ART AUCTIONS Abstract By Arzu Aysin Tekindor, Ph.D. Washington State University May 2013 Chair: Vicki McCracken Based on the field of cultural economics and economics of the visual arts, this dissertation focuses on microeconomics application to the market for fine arts. The first chapter investigates the relationship between the market value of a painting and the attributes of the painting and its artists. The artist’s attributes are represented by the primary style and primary object of the artist significantly used in the artist’s paintings. The results of the hedonic regression confirm that the market values the primary style and objects. The second chapter focuses on art museums and analyzes the effect of different types of art museums in terms of the admission policies on state residents and art visits in U.S. The study concludes that while free art is attractive for local participation, it might have a negative impact on tourist attendance. The results also show the importance of location for art museums. Finally, the third chapter analyzes premium rates and art investment by using auction sales data on impressionist and contemporary paintings. The study shows that paintings which are close to limit hammer prices tend to end up with high premium rates and low hammer prices. The art indexes that are captured from hedonic regressions show how investment on art changes with market in London and New York. Investment on impressionist and contemporary art between 2000 and 2012 in New York was relatively less risky than the market. However, investment on impressionist and contemporary art in London has no evidence for a risky investment. iv

Table of Contents Page Acknowledgements ........................................................................................................................ iii Abstract .......................................................................................................................................... iv Table of Contents .............................................................................................................................v List of Tables ............................................................................................................................... viii List of Figures ..................................................................................................................................x Dedication ...................................................................................................................................... xi CHAPTER ONE ..............................................................................................................................1 UNIQUENESS IN ART MARKET: SPECIALIZATION IN VISUAL ART–EVIDENCE FROM ART AUCTIONS ............................................................................................................................1 1. Introduction .................................................................................................................................2 2. Art Economics .............................................................................................................................4 3. The Model ...................................................................................................................................8 3.1

The Consumption Decision for Paintings – Reservation Price ..................................................... 8

3.2

Hedonic Pricing Model and Price Index ..................................................................................... 11

3.3

Data ............................................................................................................................................. 12

4. Results .......................................................................................................................................16 4.1

Artist Specific Model Results ..................................................................................................... 16

4.2

The Model with All Artists ......................................................................................................... 19

5. Conclusion .................................................................................................................................22 Bibliography ..................................................................................................................................24 CHAPTER TWO ...........................................................................................................................32 v

WHY DO WE GO TO ART MUSEUMS? IS FREE ART MORE ATTRACTIVE?...................32 1. Introduction to Art Museums ....................................................................................................33 2. How do Art Museums and Art Galleries Work? .......................................................................35 3. The Model .................................................................................................................................40 4. Empirical Model ........................................................................................................................44 5. Data ...........................................................................................................................................47 5.1

State Residents’ (Public) Participation to Art Museums............................................................. 47

5.2

Art Museums Data ...................................................................................................................... 49

5.3

States Data .................................................................................................................................. 50

6. Results .......................................................................................................................................51 6.1

State Residents’ (Public) Participation to Art Museums............................................................. 51

6.2

Art Museums’ Characteristics Effect on Art Participation ......................................................... 55

6.3

State Characteristics Effect on Art Participation ........................................................................ 58

6.4

Price Elasticity of Demand for Art Museums ............................................................................. 60

7. Conclusion .................................................................................................................................61 Bibliography ..................................................................................................................................63 Appendix ........................................................................................................................................73 CHAPTER THREE .......................................................................................................................77 ART AUCTIONS, PREMIUM RATES AND ART INVESTMENT ...........................................77 1. Introduction ...............................................................................................................................78 2. Art Auctions ..............................................................................................................................79 2.1

Christie’s and Sotheby’s Auction Houses ................................................................................... 80

2.2

Premium Rates ............................................................................................................................ 82

2.3

Collusion-Price Fixing for Auction Houses ................................................................................ 85 vi

3. Theoretical Implication .............................................................................................................86 4. Model.........................................................................................................................................87 5. Data ...........................................................................................................................................90 6. Results .......................................................................................................................................93 6.1

Premium Rates ............................................................................................................................ 93 6.1.1 Paintings Characteristics’ Effect on Premium Rates........................................................... 93 6.1.2. Changes in Premium Rates over Time Periods ................................................................... 95

6.2

Market Returns............................................................................................................................ 96

7. Conclusion .................................................................................................................................98 Bibliography ..................................................................................................................................99 Appendix ......................................................................................................................................107

vii

List of Tables CHAPTER ONE Table 1- Descriptive Statistics……………………………………………………………….…..26 Table 2- Regression Results-Artist Specific Model Results…………………………….………27 Table 3- Regression Results- All Artists Pooled………………………………………………...28 Table 4- The Marginal Effects of Style and Object on Price……...……………………….……29 CHAPTER TWO Table 1- Descriptive Statistics for Household Data……………………………………………..66 Table 2- Frequency of Art Participation for State Residents……………………….…………...66 Table 3- Descriptive Statistics for Art Museum Data…………………………………………...66 Table 4- Descriptive Statistics for State Data…………………………………………………...67 Table 5- Probit Model for Art Museum Attendance in 2008 and 2002…………………………67 Table 6- Marginal Effects- Ordered Probit Model for Art Museum Attendance in 2008……....68 Table 7- Random Effect Model for Art Museums in 2002 and 2008…………………………...69 Table 8- State Characteristics Effect on Art Participation ……………………………………...70 Appendix Table A1- Ordered Probit Model for Art Museum Attendance in 2008 and 2002……………...73 Table A2- Marginal Effects- Ordered Probit Model for Art Museum Attendance in 2002……..74 CHAPTER THREE Table 1- Buyers Premium Chart for Christie’s and Sotheby’s (Effective June 2008 until March 2013)…………………………………………………………………………………….101 Table 2- Premium rates for Christie’s Auction House in New York from 2000 to 2013……...101 Table 3- Premium rates for Sotheby’s Auction House in New York from 2000 to 2013……...102

viii

Table 4- Premium Rate Determinants and Characteristics of Impressionist and Contemporary Paintings………………………………………………………………………………………...102 Table 5- Beta Coefficient for Art Investment in London and New York...................................103 Appendix Table A1- Premium rates for Christie’s Auction House in London from 2000 to 2012….……107 Table A2- Premium rates for Sotheby’s Auction House in London from 2000 to 2012….…...107 Table A3- List of the Paintings (Two Potential Premium Rates with Final Prices)…………...108 Table A4- Price Determinants and Investment Characteristics of Impressionist and Contemporary Paintings, New York……………………………………………………………109 Table A5- Price Determinants and Investment Characteristics of Impressionist and Contemporary Paintings, London………………………………………………………………110

ix

List of Figures CHAPTER ONE Figure 1- Mean Price of Art……………………………………………………………………..30 Figure 2- S&P 500 Price Index and the Art Price Trend………………………………………..30 Figure 3- Mean Price of Artists and Artists’ Rank Trends……………………………………...31

CHAPTER TWO Figure 1- Total Appropriations to the National Endowments of Art and the State Agencies 1970-2012…………….…………………….……………………………………………..….….71 Figure 2- Price Elasticity of Demand for Art Museums- Public and Total Demand (by Selected State)................................................................................................................................71 Figure 3- Price Elasticity of Demand for Art Museums – Public and Total Demand (by Region)……………...………………………………………………………………...……..72

CHAPTER THREE Figure 1- Total Market Share (Turnover) of Christie’s and Sotheby’s………………….…….104 Figure 2- Total Market Share (Lots Sold) of Christie’s and Sotheby’s………………….…….104 Figure 3- Christie's and Sotheby's Premium Revenues (Mean) Over the Time Periods………105 Figure 4- Christie's and Sotheby's Rate of Premium Return (Mean) Over the Time Periods....105 Figure 5- Art Market Rates of Return in New York…………………………………………..........106

Figure 6- S&P500 Annualized Return….……………………………………………………...106 Appendix Figure A1- Art Market Rates of Return in London………………………………………...………111 Figure A2- FTSE100 Annualized Return…………………………………………………………..112

x

Dedication

This dissertation is dedicated to my mother and father who provided both emotional and financial support.

xi

CHAPTER ONE

UNIQUENESS IN ART MARKET: SPECIALIZATION IN VISUAL ART– EVIDENCE FROM ART AUCTIONS

Abstract This study investigates the relationship between the market value of a painting reflected by its price and the attributes of the painting and its artists. The artist’s attributes are represented by the primary style and primary object of the artist significantly used in the artist’s paintings. The theoretical model is modified from Rosen (1974) to art auctions. The results of the hedonic regression confirm that the market values the primary style and objects. We find that number of times the painting has been auctioned, popularity of the artist and gift play a significant role in the determination of the price.

Key Words: Art Valuation, Auctions, Hedonic pricing model, Reservation Price JEL Classification: Z11, D44, D11, D12

Our thanks to Goksu Kunak, an art historian who provided independent research assistance work for the data collection and coding.

1

1. Introduction A painting can be considered as a financial asset for its potential future return and as consumption good for the utility it provides from its aesthetic value and artist’s prestige. Both as an investment and as a consumption good, art markets have been the focus of economic research in recent years. This paper considers the art market in many economics concepts such as labor economics in terms of artist’s attributes, consumer behavior by considering bids in art auctions and pricing by analyzing the value of paintings. In this paper, we study how value is determined in the current art market by considering art history and focusing on some specific artists who have played significant roles in art literature such as style and object matter. A good example would be Edgar Degas who is one of the most famous impressionists and his paintings with ballerinas and dancers are well known worldwide. The paper is organized as follow. After a short introduction to the literature of research on the economics of the visual arts, a basic model of the consumption for paintings decision is developed. Art is a heterogeneous good-while there may be some characteristics that are common across items, each item is unique. Our theoretical model is developed based on a hedonic approach. The hedonic approach indicates that each good is differentiated by a set of its characteristics. Specifically here, our theoretical model assumes that the specific style of an artist and his/her specific objects, such as an apple, a woman, violin and so on, in paintings are of special importance to the reservation price which a consumer is willing to pay when other things unchanged. This paper only focuses on the tertiary art market which is the international auction market and hence our focus is only on art produced by well-known artists and previously owned by collectors, dealers and museums (Singer and Lynch, 1994).

2

In the third part of the paper, by using hedonic pricing model, we show some empirical evidence based on the data set which collected from two of the most famous auction houses, Christie’s and Sotheby’s. The data includes 15 famous artists such as Picasso, Degas and Matisse and the auctions from 1998 to 2011. The sample selected for this paper includes the artists who live in 19th or 20th centuries. These artists should be very famous in a specific style and they must be known with specific objects. The empirical model allows us to test the assumptions of the theoretical model. Of particular interest here is whether and how characteristics of paintings affect actual prices at auctions in the tertiary market. The main hypothesis considered is whether consumers value the artist’s style such as cubism or surrealism, or they just care about the name of the artist as a brand. The specific questions in this paper are: Will the value bid function increase in style and in object? How will these characteristics of painting affect the purchase? Do the bidders at auctions want to have a Picasso painting either to invest or for pleasure or do they want to have a Picasso cubist painting with abnormal shapes? Do they want to have a Degas painting or do they want to have Degas ballerina/dancer painting? We also include an artist popularity effect measured as Google hits for each artist, to analyze the effect of this characteristic on the price of an art work. This measure includes web pages related to anything about the artists such as information about paintings, biography, auctions, sales, blog comments and any historical information. Finally, last part of our paper sums up the results and makes a conclusion about how important the primary style and the primary object of the artists are. We also suggest a further study to analyze the price of the paintings.

3

2. Art Economics Art economics is considered a part of cultural economics (Blaug, 2001). Art, by definition, is a heterogeneous good. It is a durable good. Art can be defined as a private good but it can be also a public good. Knebel (2007) “There are search goods, where the quality is clear before the act of consumption and there are experience goods, where prior use is necessary in order to evaluate the quality.” The quality of some kinds of art can be defined, for example performance art. It is difficult, however, to put value on a painting by observing quality before the act of consumption. Hutter (1992) argues that the previous owners of a painting affect the value of the painting. In other words, having information about who owned the painting before and the history of the painting is an important factor for pricing. Hutter’s main point here is that the higher the art historic value, the higher the demand. Moreover, the other descriptions and classifications of art are credence good, consumer and capital goods, collector’s good, merit good and superior good. Throsby (1994) categorizes into these groups by level of market coverage and hence the competitiveness of the market. Primary market is some small galleries, provincial auctions, dealers and artists. There is relatively unlimited supply in this market and it is competitive. Secondary markets include private individuals, companies or small museums. There is monopolistic competition in these markets. Tertiary market is international auction markets where the art is produced by well-known artists. It is a monopolistic competition as well. Coffman(1991) mentions the issue of “asymmetric information” in art markets. It is much more difficult and costly to establish the value of old paintings found in an attic. Many people cannot tell the difference between paintings and photomechanical reproductions of paintings (Towne, 1969).

4

Singer and Lynch (1997) distinguish three phases of the development of a major art movement. These phases depend on the art historians; founders (who created the style), followers, artists who are following the main stream art movements (not well-known artists). In addition to Singer and Lynch’s classification of art movements, Frey (1997) came up with an exante and ex-post clarification. His point concerns prices and rate of returns. Pesando (1993) highlighted the importance of location of auction houses. He showed that Picasso’s paintings were auctioned at a higher price in New York than in London. He also found price differences among auction houses in the same city. Frey (1997) argues that the economic studies of the art market distinguish two different sources of return or utility of holding art objects. Rengers and Velthuis (2002) focus on the primary rather than the secondary market for art. They looked at gallery prices instead of auction prices. They also analyze determinants of prices rather than the rate of return on art as an investment. They claimed that all empirical studies prior to theirs are based on ordinary regression analysis to estimate determinants of prices. A weakness of this technique, according to these researches, is that it assumes that no hierarchy exists in the data (works of art, artists and galleries). For this reason, they use multilevel rather than ordinary regression analysis. Their analysis considers not only the relation between size and price for all artists but also the relation between the average price-level of an artist and his/her marginal price for additional size (centimeters). They expect that aging has a positive effect on the selling price of a piece of art. Also, they assess the effect of the residence of artists on the price of his/her painting and conclude that artists who live and work abroad are a signal of quality and charge higher price than local artists. They used both a base-line model, fixed effects model that included explanatory variables at each of the three levels (works of art,

5

artists and galleries) and random effects model in which the relation between price and size differs between artists. “No sales” takes place when the bid price is below the seller’s reserve price. Ekelund, Ressler and Watson (1998) say “Auction houses usually tell potential sellers that the reserve price is 80 of the low presale price estimate. This may indicate that the reserve price is easily determined by observing the low price estimate.” These researchers analyzed whether there is a bias in the auction house estimates in Latin-American Art. They found that the size of the painting has no predictable effect on price. Ashenfelter and Graddy (2003) have conducted theoretical and empirical research on strategic behavior in auctions. In their research, they also discuss the mechanics of art auctions. One of the rules of an auction is that auction houses charge commissions to buyers and sellers. The total sale price to the buyer is the sum of the hammer price (the final price for the item in the auction) and the buyer’s premium (around 10-17.5 percent of hammer price). Sellers also pay commission to the auction houses (10 percent of hammer price). Moreover, Ashenfelter and Graddy also analyze how art asset prices over time (time-series movement). They use a hedonic model with repeat-sales data and include a dummy variable for each painting. Using repeat-sales data on impressionist art, allowed them to determine whether there were systematic differences in prices and estimates for paintings that came to auction and did not sell during their first appearance at auction but sold during their second appearance at auction vs. those that came to auction and sold during their first appearance and were resold again during their second appearance at auction. They found that the second appearance value of a painting was unsold at auction is negatively influenced by not being sold during its first appearance. They also discuss “secret reserve prices”. The reason for using secret reserve prices is to deter collusion.

6

Knebel (2007) used Google to obtain information about the amount of media attention attracted by an artist and how the public response to this attention. One of the variables in his data is “count of total hits for artist at Google for search in English”. Examples of hits that Knebel used include biographical pages, information about exhibitions, information about the latest sales and excerpts from printed articles. Knebel claims that his work was the first time that popularity data was used in this way in with the fine art market and artists. Using Google search variable in the data, he looked for an evidence for the existence of superstardom. Knebel’s results however do not provide sufficient evidence to support the existence of superstardom in the visual arts. We also use Google hits for the popularity of the artist and see how the popularity affects the price of the painting. In the work by Hutter, Knebel, Pietzner and Schafer (2007), two major hypotheses were considered. Hypothesis was that prices of the paintings from living artists increase steadily over their lifetime; however, the auction prices for works by the same artist do not show such a pattern. The second Hypothesis was that the aggregate price level in dealer markets is higher than in auction markets for works by the same artist and the dealer price index follows upward movements but not downward movements of the auction price index. They found that prices increase with age in a nonlinear pattern. Moreover, auction prices are lower than dealer prices and the price increases due to age are lower in auction markets. Our paper contributes to the existing art economics literature in our focus on the specific characteristics of a painting that is the style and object in the painting in order to identify their effects on price. In addition, we also find significant results about gifted paintings and paintings which have been auctioned several times in the past.

7

3. The Model 3.1

The Consumption Decision for Paintings – Reservation Price

A painting can be described by a number of characteristics which all contribute to the value attached to the complete art. The model developed for this research focuses on a hedonic hypothesis which considers that each good is differentiated by the set of all of its characteristics,

zi (i  1,2,3...n) . zi is defined as the amount of ith characteristic in each good. It takes a value of 0 or 1 for the existence of a characteristic or can be a continuous variable for ith characteristic. This model depends on product differentiation and addresses the questions which concern the behavior of style and object in paintings at auctions are of special importance to the reservation price a consumer is willing to pay at a given income and utility. The specific question is: Will value bid function increase in style and object? The first assumption of the model is that buyers purchase only one painting of an artist with a particular value of set z . The strictly quasi-concave utility function is represented as

U (k , z1, z2 ...., zn ) where k is consumption of a numeraire non-painting good by assuming that price of all other goods is normalized to 1, z is painting. It follows that U (k , z ) 0 z

and

U (k , z ) 0 k

(1)

i.e, the utility function is increasing in z and k. The budget constraint is represented as m  k  p( z )  

(2)

where m is income and  is buyer’s premium. Auction houses charge a fee known as a Buyer's Premium. This is added to the winning bid (hammer price) and is part of the total purchase price and hence part of the budget constraint.

8

The buyer will reach utility u (uniform utility level) and the payment or the reservation demand price,  ( z1 , z2 ,..., zn ; u, m) , under the condition of

U (m   ( )   , z1 , z2 ,..., zn )  u

(3)

where  ( z; u, m) is consumer bid function which is the reservation price a bidder is willing to pay for values of the set of all characteristics of painting, zi . If the bidder wins the auction and buys the painting, then z  1 and the utility is U (m   ( )   ,1) . If the bidder does not buy the painting then the utility is U (m,0) . The bidder buys the painting if U (m   ( )   ,1)  U (m,0) .

After setting up the utility for consumers for the general characteristics of painting, specific utility function can be defined with z1 and z2 for object and style, respectively. Then, the reservation price a bidder is willing to pay for object and style is  ( z1 , z2 ; u, m) when the other components of z are held constant. The utility maximizing painting choice can be determined by maximization of

U

with respect to all painting characteristics.

The buyer's bid price function is derived by maximizing (3).   U z1  U k z1  U k   z1  0 , z1

 z (1   ) 

U z1

  U z2  U k z2  U k   z2  0, z2

 z (1   ) 

U z2

  1  U ku  U k  u  0, u

u  

  1   m   m  0, m

0

(4)

0

(5)

1 0 U k (1   )

(6)

1 0   1

(7)

1

2

m 

9

Uk

Uk

Note that  z1 (1   ) and  z2 (1   ) are marginal rate of substitution between object, style, respectively, and the numeraire which is money for all other goods. If the marginal utility of object is greater than the marginal utility of style, then consumers’ utility is greater when the painting includes artist’s object rather than the artist’s style. We consider that the object is more specific with the artist than style. Owning a painting with an object which is known with a specific artist is more recognizable than the style. Equation (4) and (5) show that the bid price function (also known as the reservation demand price) is increasing in style and object. It means that the buyer is willing to pay more for the painting which includes artist’s object and/or style. Also, equation (4) and (5) imply that

 z   z when U z  U z . Therefore, the buyer is willing to pay more for object than style. 1

2

1

2

Alternatively, additional income increases maximum attainable utility. Equation (7) shows that the reservation demand price is increasing in income but at a decreasing rate with the buyer premium. Finally, for the maximum utility, reservation price should equal to p(z) and  . Then, a general hedonic model can be written as:

 ( z * ; u * , m)  p( z * )   ( ( z * ; u * , m))  z ( z1* , z2 ; u * , m)  p( z1* )   1

 z ( z1 , z * ; u * , m)  p( z2* )   2

2

Since  ( z; u, m) is the consumer’s willingness to pay and p( z)   ( ( z; u, m)) is the minimum hammer

price

he

must

pay

for

the

10

painting,

utility

is

maximized

when

 ( z* ; u* , m)  p( z* )   ( ( z* ; u* , m)) where z * and u * are optimum quantities and  is the buyer’s

premium.1 3.2

Hedonic Pricing Model and Price Index

The evolution of the art auction market has been studied in a number of previous papers. Two approaches have been used for analyzing art markets: repeat sale regressions and hedonic price regressions. According to Chanel et al (1996), hedonic models generally yield more reliable estimates than repeat sales. Chanel suggests that price indices of paintings should be based on regressions using the full set of sales, and not resale only. They show that the standard deviations from the regression are an accurate measure of the value with which indices over time are estimated by a hedonic regression. The hedonic price function reveals information about the structure of the preferences of consumers and shows the relationship between a product’s price and its characteristics. The hedonic price function demonstrates how much a consumer has to pay to obtain different bundles of characteristics. The most common econometric model used for art auction is hedonic price regressions. Onofri (2009) analyzes the price of Old Master Paintings. Ashendfelter and Graddy (2003) estimate rates of returns and price indices by both using the hedonic model and repeat-sale model. Czujack (1997) examines the market of Picasso paintings sold at auction between 1963 and 1994 by using hedonic approach. Chanel (1995) and Gerard-Varet (1995) use hedonic models to construct price indexes of art.

1

See Rosen, Sherwin (1974), “Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competation”, The Journal of Political Economy, Vol 82, No.1, page 39. .

11

A hedonic function with price as the dependent variable was specified, and estimated using an equation of the following form: n

t

i 1

 1

ln Prt   0    i zir   t wrt  urt where Prt is the price of the painting r sold at time t , zir the i th characteristics of the painting

r . i does not necessarily depend on t the year in which the painting is sold. wrt is a dummy variable taking the value of 1 if the painting is sold in year t and 0 otherwise. We use the logarithm of the price as a dependent variable and hence the estimated coefficient of a dummy variable allows an assessment of the estimated percentage difference in price due to the painting having the particular characteristic described by the dummy.2 3.3

Data

The data set consists of prices and characteristic of 1103 paintings by worldwide major artists, auctioned and sold by Sotheby’s and Christie’s in New York, London and Paris between 1998 and 2011. Prices of paintings are in US dollars and include what is known as the buyer’s premium. The information about paintings is made available by Sotheby’s and Christie’s online catalogue before and after the auction. These catalogues contains “estimated low” and “estimated high” prices which are produced by art experts such as art historians in the auction houses’ departments. The data do not include any “bought in” (unsold) paintings and thus reflect only actual sales. If the hammer price is below the reserve price, the sale does not occur. The auction houses do not announce the items that were withdrawn, passed, or unsold in the auction. The data selection process focuses on famous artists and included works of the following artists; Picasso, Dali, Munch, Klimt, Degas, Matisse, Cassatta, Miro, Kooning, Picebia, Chagall, 2

Box-Cox test suggests the logarithmic transformation of the dependent variable.

12

Wassily, Hassam, Marc and Magritte. The explanatory variables considered pertain either to the characteristics of the artists or the specific painting or auction, or to time, and are measured as follows. Artist: We use 15 different dummy variables, one for each artist to indicate observations in the sample for that artist, all artists included lived in the 19th or 20th centuries and was very famous in a specific style and they must be known with specific objects. The data also include the age of the artist when the painting was executed. This age range is from 16 to 95. Painting: We have 26 different variables define the painting characteristics, such as medium, number of previous owners, size, executed date, signature, exhibition, gift to first owner, estate of the artist, style of the painting such as realism, cubism, fauvism, expressionism, impressionism and abstract. We use a dummy variable to indicate whether a specific painting represented that artist’s primary style. For example, Salvador Dali is well known with surrealism, Degas, Cassatt and Hassam with impressionism, Munch with expressionism and Chagall with symbolism. Picasso is famous with cubism. “Style” is a dummy variable taking the value 1, if the painting was painted with the artist’s primary style. Likewise a dummy variable is used to indicate whether the object in a painting is well known with the artist. For instance, Klimt is well known with using erotic women in his paintings. Magritte used apples, pipes and some other object in his paintings. Cassatt mostly focused on mother and child paintings. Dali is known with melting clocks, Gala (his wife) and sticks. Much inspired by death, illness (sick people) and darkness. Hassam’s best known paintings are the “flag” paintings. “Object” is a dummy assuming the value of 1 it the paintings has the artist’s primary object or objects.

13

The dummy variable, style, shows if the painting was painted by the artist’s primary style. Also the dummy variable, object shows if the painting has the artist’s primary object. We also have an interaction term which indicates if the painting has both style and object of the artist. This variable will be introduced to our model with all artists (Table 4). Figure 1 shows the mean prices for the paintings which are categorized by having the artist’s major style, object, both and none.

Auction houses: The data set was collected from two of the most famous auction houses, Christie’s and Sotheby’s. The type of the auction is “English” or “ascending price” auctions. Bidding starts with a low price and is raised as progressively higher bids until the bidding stops. The final price is called “hammer price”. If the bidding does not reach its (secret) reserve price, the item will go “unsold”. The unsold item is rarely bought by the auction house. Auction houses’ income comes from commissions from both sellers and buyers. Christie’s and Sotheby’s increased their buyer premium rate simultaneously. According to Ashenfelter & Grebby (2002), in some cases, the seller may pay no commission and even be guaranteed a minimum sale price. This may be a consequence of competition between auction houses. We use a dummy variable to indicate which auction house a painting was associated into and a dummy variable to indicate which city; London, Paris and New York, where to painting was sold. Time variable: Our data was selected over 13 years, 1998- 2011. Because of market and general economic changes over this time period, we introduce a dummy variable for each year of sale. We consider only sale year. An additional time related variable, recession was captured to

14

indicate years that were considered as recessionary years in U.S., U.K and France. 3This variable shows the effect of recession on the price. Table 1 shows descriptive statistics for the data. The most expensive painting in the data belongs to Picasso. It was sold for $104,000,000 in 2004 at Sotheby’s auction house. The least expensive painting is from Cassatt with the price of $6,000. The greatest negative difference between the auction price and estimated high price released is a Matisse painting in 2010 with the difference being -14,900,000. This indicates that the highest expert’s valuation for this painting was almost $15 million more than resulting price. It seems that the experts overestimated for this painting. The greatest difference between the lowest estimated value for one of Matisse’s painting and its actual price was almost $5 million in 2010 in Christie’s auction house. This shows that this painting was sold $5 million less than what it was expected to be. Matisse’s paintings are more surprising than those of other artists’ paintings in the data. Another example supporting this conjecture is that the greatest positive difference between the price and the highest estimated price is $29,900,000 again for Matisse’s painting in 2009 at Christie’s’ auction house. Since the data is collected from the international painting market, we use the S&P 500 price index. The relationship between S&P 500 price index and the price trend of art is demonstrated in Figure 2. While the price trend for art increases over time, S&P 500 price index’s increase is not significant.

3

The dummy variable recession for U.S. is defined for time periods; March 2001- November 2001 and December 2007-June 2009. The dummy variable recession for U.K. is defined for time periods; April 2008-May 2009 and October 2010-December 2010. The dummy variable recession for France is for May 2008- December 2009.

15

4. Results 4.1

Artist Specific Model Results

First, we run the model without time variables and a style-object interaction. The logarithm of the auction price is adjusted by S&P 500 index and used as a dependent variable in all models. Table 2 shows the results for 8 artists; Picasso, Dali, Magritte, Klimt, Matisse, Kooning, Chagall and Degas. These artists were selected because of being relatively most famous ones and availability of data. Pablo Picasso: Picasso is one of the most well-known artists in the world and a cofounder of Cubist movement. Google also provides support for this: Picasso is the most popular artist in our data in terms of the number of Google hits. He had eight different working periods, however we consider Cubism as his major style. Czujack (1997) examined Picasso’s paintings and our results are similar to Czujack. He found that, publication and provenance do not matter for the price. He expected “Style Picasso” paintings would be more expensive than other Picasso paintings. However, they concluded that this was not true. Our model results are similar for Picasso’s painting which sold at auctions for the years between 1998 and 2011. An addition, we also show that the object does matter for the Picasso’s paintings. Moreover, our model suggests that if a painting was a gift from Picasso to the first owner of the paintings, then the price at subsequent sells was negatively impacted. Salvador Dali: The greatest known Surrealist artist is the world famous Salvador Dali.

4

He is the most famous representative of Surrealist movement. We found that paintings that are gifts from Dali had statistically significant lower prices in later auctions, similar to the results for Picasso. The results show that style for Dali’s paintings is very important for price. Style has a 4

http://www.surrealism.org/, 11/30/2011

16

positive impact on price. Although the coefficient for object is positive, it is not statistically significant. Rene Magritte: Magritte’s earliest oil paintings were impressionistic in style. He tried Cubism and became famous with Surrealistic style. He is mostly well known with the objects he used in his paintings such as pipes, apples and clouds. Our results do confirm that for Magritte it his object, and not style, that was a significant effect on the painting price. Gustov Klimt: Klimt’s immensely erotic portraits and sexually-charged sketches are refined and remain among the most recognized works of art in the world.5 He has very unique style. “In 1897 Gustav Klimt founded with other artists the Vienna Secession and became its first president. By that time Klimt had developed his own and characteristic style, which should became the trademark of the movement. Like impressionism, also art nouveau was an international revolt against the traditional academic art style.”6 We found that style has a positive effect while object has negative effect on the price of Klimt’s paintings. Eroticism as the object in his paintings and the social moral norms might partially explain these results. Henri Matisse: The leader of Fauvism was Henri Matisse. Matisse’s paintings are recognizable in terms of colors and patterns. Fauvist paintings have usually bright, row and unusual colors. He usually painted women figures. “His varied subjects comprised landscape, still life, portraiture, domestic and studio interiors, and particularly focused on the female figure.”7

5

http://www.iklimt.com/, 11/30/2011 http://www.artelino.com/articles/gustav_klimt.asp, 11/30/2011 7 http://www.metmuseum.org/toah/hd/mati/hd_mati.htm, 11/30/2011 6

17

Style has positive effect on the price of Matisse’s paintings in our research. As we mentioned before, Matisse’s paintings are the most extraordinary paintings in our data in terms of unexpected price results. The results show that size of the painting significantly impacts price. Also, Matisse’s signature on the painting has a positive significant impact on the auction price. Willem de Kooning: Kooning is an abstract expressionist. He is one of the best representatives of this style. His women series are on the list of the most expensive paintings in the world.

8

He usually painted on very big canvases. “Kooning’s paintings were not only

shocking for their perceived hostility towards women, which is subject to debate, but also because de Kooning focused on the human figure at a time when art world dogma praised abstraction almost exclusively.”9 Both style and object have positive and significant effects on the price of Kooning’s paintings. That is, Kooning’s women figures and his abstract expressionist style both increase the price of the painting. There is a significant effect of size. Marc Chagall: Chagall created the art forms of Cubism, Symbolism and Fauvism. Later on, the influence of Fauvism made him to rise to Surrealism.10 He focused on village life and mostly goats, bride-groom, cows and Surrealism. We found that both object and style have positive effects on the price of Chagall’s paintings. Edgar Degas: There is no doubt that Degas is one of the most famous impressionists. His ballerina/dancer paintings and also sculptures are exhibited all around the world in very famous museums. “He is regarded as one of the founders of impressionism although he rejected the

8

http://www.technology.am/the-30-most-expensive-paintings-of-all-time-141346.html 12/01/2011 http://edu.warhol.org/app_dekooning.html, 12/01/2011 10 Lewis, Michael J. "Whatever Happened to Marc Chagall?", Commentary, October, 2008 pgs. 36–37 9

18

term, and preferred to be called a realist.”11 “He is especially identified with the subject of the dance, and over half his works depict dancers.” 12 We found that the object has a positive significant effect on price for Degas’ paintings. Surprisingly, we found that if the painting was Degas’s estate, this decreases the price of the painting. Those paintings in Degas’s studio were mostly sketches. Also, like other artists, the size of Degas’ paintings has impact on the price. 4.2

The Model with All Artists

In this paper, a hedonic model is estimated to analyze the art market. We introduce the interaction term for style and object. Results from OLS estimation are displayed in Table 3 and Table 4. According to the Table 3, the resulting equation explains 65% (R2=0.65) of the total variation in painting prices. Painting: Oil on canvas is the most expensive combination. To be able to compare the medium of paintings, the dummy variable representing watercolor is used as a based category in our regression model (i.e. perfect multicollinearity). For this reason, the estimated coefficients for the type of surface measures the impact of the specific surface relative to watercolors. The results show that pencil is statistically not preferred to watercolor paintings. In other words, the media watercolor, drawings and other types such as crayon (“other type of medium” variable) generate lower price than in oil. We found that prices are increasing with surface. The dimension of painting has an effect on the resulting price but this effect is very small (0.008%). The sign of the variable “dimension square” shows that the size effect on price weakens the larger the painting is.

11 12

Gordon, Robert; Forge, Andrew (1988), ” Degas” New York: Harry N. Abrams. ISBN 0-8109-1142-6 www.edgar-degas.org/

19

We tested that number of owners who owned the painting before, has positive effect on the price. This increases the price by 8.5%. However, estate of the artist has no statistically significant effect on price. The number of auctions of painting until its last auction date has a significant effect on price. In particular, one unit increase in number of auctions results in a 15% decrease in price, all else constant. Moreover, we found that if the painting was a gift to the first owner or if it was an exchange between two artists, then the resulting price was around 26% lower than otherwise. This shows that buyers think that paintings that were given away by its artist are not as precious as the paintings which were sold to its first owner. Age of an artist when the painting was executed has a positive effect on the selling price but with a decreasing rate. The effect of age is tempered somewhat by negative coefficient for age squared. The result shows that every year increase in age of an artist adds %5.2 more onto the price. Signature, Literature and Exhibit are indicators of prestige. Our regression shows that if the painting has these prestige characteristics, this increases the price of the painting. For example, the signature of a painting has a positive effect of 50% on the resulting price. Also, we compare realist paintings with other styles; impressionist, cubist, fauvist, expressionist, abstract surrealist paintings and undefined paintings which have not specific style. Model 2 in Table 3 also shows the ceteris paribus comparison. 13 The interaction variable of object and style measure the differences in change of price for the object when the painting has also artist’s primary style. With a negative influence of 0.29, it is shown in Table 4 that having artist’s primary object and style together has less influence on the resulting price. F-test (significant at the 0.05 confidence level) for the marginal effects of 13

Picasso is also treated as a based category because of collinearity. Since the artist has a single value for the Google hits, hence the estimating eliminates one more of the artists.

20

style and object on price shows that coefficient of style and the coefficient of object are not the same. Therefore, we conclude that consumers are willing to pay more for the object than for the style. That is, the effect of the object is statistically greater than the effect of style on the price. Regarding to our theoretical model, our hypothesis is accepted. Auction Houses: We found that sale at Sotheby’s auction house has a positive effect on price. Also, price obtained in New York is more expensive than sale in Paris. Time: The economic cycle is an important explanatory factor of demand for art works. Our estimation suggests that recession has a relevant effect upon the total purchase price of a painting. Not surprisingly, recession decreases the prices of works of art. In other words, it decreases demand on art work and causes a price drop. We found that during the recession in the U.S., U.K and France the price of the painting decreases around 40% all else constant. Artist: Model 1 in Table 3 also shows the ceteris paribus comparison of the impact of artist on price, in the coefficient indicatory the comparison of Mary Cassatt (based category) and other artists. Picasso’s paintings (132%), Matisse’s paintings (147%), Klimt’s paintings (138%) and Degas’s paintings (73%) generate higher price than Cassatt’s paintings. Popularity (Google hits): Google is the dominant search engine in the United States market, with a market share of 65.6%.14 The total number of hits registered by Google has been used by many academic researchers. According to Knebel (2007), the information on the number of Google hits per search word is used as an approximate measure of the popularity of each artist. We also use Google hits for artists’ full names. Hits on Google show the web pages related with the artists. The information in these web sites can be anything about the artists such as

14

"comScore Releases November 2009 U.S. Search Engine Rankings". December 16, 2006. Retrieved July 5, 2010.

21

paintings, biography, auctions, sales, blog comments and any historical information. Figure 3 shows the artists in our data and their prices and rank trends (number of Google hits). Finally, the variable Popularity was introduced to our regression in Model 2 (Table 3). We use the logarithm of Google hits as an independent variable. A positive sign for this term would show that popularity of an artist is important for the resulting price at auction market for famous artists. This conjecture is confirmed with highest significance levels (1%) with a factor of 0.38 which means that 1% increase in popularity (Google hits) results in a price increase of 38%. By including the popularity of the artists in Model 2, all coefficients slightly changed. The resulting equation with the Google hits explains 64% (AdjR2=0.64) of the total variation in painting prices. As in our first model, in Model 2, the effect of object is statistically greater than the effect of style on the price. Our hypothesis is also accepted when the popularity effect is added into the model.

5. Conclusion This paper focuses on artists’ specific styles and objects which made them different and recognizable in the art world and analyzes effects of these characteristics on price. We have tested for the significance of several factors which are expected to influence the price of paintings of some famous artists who lived in 19th and 20th centuries. We have used a relatively homogenous cross-sectional data consisting of modern art paintings. We found that primary style paintings of an artist are more expensive than an otherwise comparable painting by the same style. Also, presence of primary object of an artist increases the price of the painting. The hypothesis which states that consumers’ utility is more when the painting includes artist’s object 22

rather than the artist’s style. That is, the buyer is willing to pay more for object than style. By using hedonic regression, we found that effect of object is statistically greater than the effect of style on the price and also found that having artist’s primary object and style together has less influence on the resulting price. In addition to effects of style and object, we also found significant results about gifted paintings and paintings which has been auctioned several times in the past. The number of times the painting has been auctioned has a negative significant effect on price. The more the painting has been sold at auctions, the less the price was sold for. This finding might be demonstrating perceived attainability. If a painting was on the market several times before, this shows that the painting is easy to attain and this decreases the price of the painting. Whether the painting was a gift to the first owner or if it was an exchange between two artists, this affects the price in a negative way. In the bidder’s point of view, the gift shows the artist’s valuation on the painting. It is a signal of preciousness which was assigned by artist for the gifted painting. Our research does represent the importance of style and object for price. We show that having specific object and style increase the price. The regressions were also estimated for some artists. As it was expected, we found that style for Dali’s painting and objects for Magritte’s and Degas’ paintings have significant impacts on price. On the other hand, this might also depend on what the object is. For example, the effect of object (e.g. erotic figures) on the price of Klimt’s paintings is negative. Social moral effect might be the reason of this result.

23

Bibliography Ashenfelter, Orley and Kathryn Graddy (2002), “Art Auctions: A survey of Empirical Studies”, Working Paper 8997, National Bureau of Economic Research. Ashenfelter, Orley and Kathryn Graddy (2003) “Auctions and the Price of Art”, Journal of Economic Literature, Vol. 41(3), 763-787. Blaug, Mark. (2001), “Where Are We Now on Cultural Economics?”, Journal of Economic Surveys, Vol. 15, 123-142. Brueckner, Jan K. and Peter F. Colwell (1983), “A Spatial Model of Housing Attributes: Theory and Evidence”, Land Economics, Vol.59, No.1, 58-69. Campos, Nauro F. and Renata L. Barbosa (2009), “Paintings and Numbers: an Econometric Investigation of Sales Rates, Prices, and Return in Latin American Art Auctions”, Oxford University Economic Papers, Vol. 61, 28-51. Chanel, Oliver, Louis-André Gérard-Varlet and Victor Ginsburg (1996), “The Relevance of Hedonic Price Indices: the Case of Paintings.” Journal of Cultural Economics, Vol. 20, No. 1, 124. Coffman, Richard B. (1991) “Art Investment and Asymmetrical Information”, Journal of Cultural Economics, Vol. 15, 83-94. Czujack, Corinna (1997) “Picasso Paintings at Auction, 1993-1994”, Journal of Cultural Economics, Vol. 21, 229-247. Ekelund, Robert A. Jr, Ressler W. Rand and John K. Watson, (1998) “Estimates, Bias and “No Sales” in Latin-American, Art Auctions, 1977–1996”, Journal of Cultural Economics, Vol.22, 33-42 Frey, Bruno S. (1997), “Art Markets and Economics: Introduction”, Journal of Cultural Economics, Vol. 21, 165-173. Gordon, Robert and Andrew Forge (1988), “Degas”, New York: Harry N. Abrams, ISBN 08109-1142-6. Hutter, Michael (1992), “Die bildende Kunst und ihre Wirtschaft. Kulturmanagement Kulturökonomik.” Hagen, Fernuniversität Hagen. Hutter, Michael, Christian Knebel, Gunnar Pietzner and Maren Schäfer (2007) “Two Games in Town. A Comparison of Dealer and Auction Prices in Contemporary Visual Arts Markets.” Journal of Cultural Economics, Vol. 31, Issue 4, 247-261.

24

Knebel, Christian (2007), “Anomalies in Fine Art Markets-Three Examples of an Imperfect Market for Perfect Goods”, http://www.scribd.com/doc/48558040/Anomalies-in-Fine-ArtMarkets-Three-Examples-of-an-Imperfect-Market-for-Perfect-Goods Onofri, Laura (2009), “Old Master Paintings, Export Veto and Price Formation: an Empirical Study”, European Journal of Law and Economics, Vol. 28, 149-161. Pesando, James. E. (1993), “Art as an Investment: The Market for Modern Prints”, American Economic Review, Vol. 83, 1075-1090. Pommerehne, Werner W. and Lars P. Fars (1997) “The Impact of Museum Purchase on the Auction Prices of Paintings.” Journal of Cultural Economics, Vol. 21, 249-271. Rengers, Merijn and Olav Velthuis (2002) “Determinants of Prices for Contemporary Art in Dutch Galleries, 1992–1998” Journal of Cultural Economics, Vol. 26, 1–28. Rosen, Sherwin (1974), “Hedonic Prices and Implicit Markets: Product Differentiation in Pure Competation”, The Journal of Political Economy, Vol. 82, No.1, 34-55. Singer, Leslie and Gary A. Lynch (1994), “Public Choice in the Tertiary Art Market”, Journal of Cultural Economics, Vol. 18, 199-216. Singer, Leslie. P. and Gary A. Lynch, (1997), “Are Multiple Art Markets Rational?”, Journal of Cultural Economics, Vol. 21, 197-218. Stanley, Dick, Judy Rogers, Sandra Smeltzer and Luc Perron (2000) “Win, Place or Show: Gauging the Economic Success or the Renoir and Barnes Art Exhibits”, Journal of Cultural Economics, Vol. 24, 243-25. Throsby, David. (1994) “The Production and Consumption of the Arts: A View of Cultural Economics.” Journal of Economic Literature, Vol. 32, 1-29. Towne, Morgan (1969), “Treasures in Truck & Trash.” Post Washington: Kennikat Press. Ulibarri, Carlos A. (2005), “Bayesian Learning from Arts Goods?- A Comment”, Journal of Cultural Economics, Vol. 29, 137-141.

25

Table 1-Descriptive Statistics Variable Auction Price Adjusted Price (S&P 500) Estimated Low Price Estimated High Price Auction Price minus Estimated Low Price Auction Price minus Estimated High price Executed year Age of painting Provenance Number of auctions Artist's age*

Obs 1103 1103 1095 1095

Mean 2,395,895 2,594,039 1,339,635 1,885,045

Std. Dev. 6,327,637 6,890,730 2,708,050 3,783,837

Min 6,000 6,107 2,845 4,268

Max 104,000,000 121,000,000 24,000,000 30,000,000

1094

749,783

2,236,685

-4,909,500

31,000,000

1090 1055 1050 1051 1054 1055

502,254 1932 75 4 0.5 53

1,918,587 29 29 2 1 17

-14,900,000 1855 16 1 0 16

23,900,000 1987 154 14 6 96

*Age of the artist when the painting was executed.

26

Table 2-Regression Results-Artist Specific Model Results ln(Auction Price) Painting

27

Oil Pencil Other type medium Style Object Paingting’s age (Paingting’s age)2 Number of Owners Dimension (Dimension )2 Number of Auctions Estate of the artist Gift to the first owner Literature Signed Exhibited

Picasso

Dali

Magritte

1.124** -0.662 0.101 0.205 0.392* 0.103* -0.001** 0.096 0.000*** -0.000** -0.303* 0.176 -0.759* 0.117 0.483* 0.182

0.77*** -0.213 0.537 0.815*** 0.237 0.068 -0.001 -0.130* 0.000*** -0.000*** 0.210 0.831 -1.351*** 0.480* 0.703* 0.331*

0.666* -0.359 -0.552 -0.198 1.238*** 0.090 0.001 0.127*** -0.000 0.000* -0.137 0.3100 -0.401 -0.051 1.685*** 0.248 0.255*** -0.001

Klimt

Matisse

Kooning

Chagall

Degas

1.976*** -0.612 (omitted) 1.385*** -1.258*** 0.116 -0.001 0.026 -0.000 0.000 -0.007 0.275 (omitted) 0.068 0.290 0.249

0.693** -0.143 (omitted) .497** 0.234 -0.117 0.001 0.045 0.000*** -0.00*** 0.315* 0.306 -0.335 0.405 0.909** 0.371

0.208 0.610 0.415 0.901** 0.823*** 0.080 0.000 0.298*** 0.000*** -0.000*** -0.304* 0.239 -0.374 0.361 0.326 0.403

1.465*** -0.642* 0.768** 0.732*** 0.481*** 0.149* -0.001 0.026 0.000*** -0.000** 0.048 -0.118 -0.357 0.241 -0.097 0.081

(omitted) -0.328 0.108 0.461 0.627** 0.232 -0.001 0.097 0.001*** -0.000* -0.132 -0.924*** -1.205 0.657 0.345 0.001

0.170 -0.002

0.244 -0.002

-0.096 0.001

-0.051 0.001

0.156 -0.002

-0.187 0.741* -0.094 6.746 0.612 93

0.258 (omitted) -0.021 7.398 0.717 81

0.443** -0.466 0.188 3.557 0.880 78

-0.102 -0.181 0.146 -8.192 0.500 104

Artist -0.0219 Artist’s Age -0.213** 0.000 (Artist’s Age)2 0.001* Auction House -0.268 New York -0.071 -0.873 Paris 1.391** Auction House(Christie’s=1) -0.670*** -0.575*** 7.186 ** 18.88*** Constant 2 0.720 0.708 R 104 114 Number of observation *** significant at the 0.01 level , ** significant at the 0.05 level,

-0.051 -0.409 -2.788*** (omitted) -0.114 -0.368* -6.703** 4.206 0.804 0.853 97 71 *significant at the 0.1 level

Table 3- Regression Results- All Artists Pooled Model 1 ln(Auction Price) Painting (watercolor based category) Oil Pencil Other type of medium Style Object Object & Style Painting’s age (Painting’s age)2 Number of Owners Dimension (cmSquare) (Dimension (cmSquare)) 2 Number of Auctions Estate of the artist Gift to the first owner Literature Signed Exhibited Auction House (London based category) New York Paris Auction House(Christie’s=1) Artist Artist’s Age (Artist’s Age)2

Picasso Dali Munch Klimt Degas Magritte Kooning Picebia Miro Chagall Matisse Wassily Hassam Marc Popularity (ln Google hits) Painting Style (realism based category) impressionist expressionist surrealist

Coefficient

Model 2 Standard Error

Coefficient

Standard Error

1.035*** -0.491*** 0.521*** 0.349** 0.771*** -0.286* 0.003 0 0.085*** 0.000*** -0.000*** -0.150*** -0.074 -0.295** 0.455*** 0.504*** 0.383***

0.137 0.16 0.149 0.144 0.147 0.167 0.006 0 0.021 0 0 0.052 0.102 0.145 0.089 0.103 0.080

1.034*** -0.490*** 0.521*** 0.353** 0.772*** -0.286* 0.003 0 0.085*** 0.000*** -0.000*** -0.151*** -0.073 -0.295** 0.455*** 0.504*** 0.383***

0.137 0.160 0.150 0.145 0.147 0.167 0.006 0 0.021 0 0 0.052 0.102 0.145 0.090 0.104 0.080

0.210*** 0.138 -0.179**

0.080 0.185 0.070

0.201** 0.130 -0.180***

0.080 0.185 0.070

0.052*** -0.001***

0.014 0.000

0.014 0.000

(Cassatt based category) 1.322*** 0.116 0.756** 1.376*** 0.731*** 0.297 0.910* -0.656* 0.754* 0.399 1.467*** 0.917*** 0.783** 0.061

0.052*** -0.001*** (Cassatt and Picasso based category)

0.331 0.432 0.31 0.312 0.229 0.403 0.500 0.347 0.404 0.362 0.291 0.315 0.345 0.426

-0.318 0.457* 0.736*** 0.363* 0.387 0.971* -0.324 0.585 0.012 1.043*** 0.692** 1.181*** -0.008 .388***

0.349 0.269 0.246 0.218 0.422 0.512 0.414 0.372 0.288 0.221 0.281 0.417 0.417 0.097

0.305* 0.383* 0.549***

0. 170 0. 200 0. 201

0.304* 0.384* 0.544***

0.170 0.200 0.201

28

cubist fauvist abstract undefined symbolist

0.559** 0.596*** -0.417 0.057 -0.522* -0.360**

0. 219 0. 223 0. 254 0. 157 0. 275 0.142

0.556** 0.592*** -0.420* 0.056 -0.526* -.342**

0.219 0.223 0.254 0.158 0.275 0.143

-0.293 -0.622** -0.677** -0.556*** -0.514** -0.272 -0.254 -0.393* -0.043 -0.393* 0.277 0.373 -0.072 8.580*** 0.64 1012

0.296 0.308 0.256 0.236 0.238 0.264 0.230 0.220 0.210 0.204 0.227 0.227 0. 207 1.035

-0.289 -0.618** -0.676*** -0.555** -0.509** -0.268 -0.249 -0.388* -0.038 -0.387* 0.269 0.371 -0.066 3.050 0.64 1012

0.296 0.308 0.257 0.236 0.238 0.264 0.230 0.220 0.210 0.204 0.228 0.227 0.207 2.259

During recession Year (2011 based category) 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Constant Adj R2 Number of observation

*** significant at the 0.01 level , ** significant at the 0.05 level, *significant at the 0.1 level

Table 4-The Marginal Effects of Style and Object on Price Model 1

Model 2

Object Style

0

1

0

1

0

0

0.771

0

0.772

1

0.349

0.834

0.353

0.839

29

Figure 1-Mean Price of Art 3,500,000 3,000,000 2,500,000 2,000,000 1,500,000 1,000,000 500,000 0 Only Primary Style Primary Object & Only Primary Object (No Primary Object) Primary Style

Neither

Number of observation:1103

Figure 2- S&P 500 Price Index and the Art Price Trend Art Price (S&P Adjusted)

S&P 500 index

4,500,000

1,600

4,000,000

1,400

3,500,000

1,200 S&P500 Adjusted price

3,000,000

1,000

2,500,000

800

2,000,000 1,500,000 1,000,000

400

500,000

200

0 1995

S&P500 index (monthly)

600

2000

2005

2010

30

0 2015

Linear (S&P500 Adjusted price) Linear (S&P500 index (monthly))

Figure 3-Mean Price of Artists and Artists’ Rank Trends

7,000,000

50,000,000

6,000,000

40,000,000

5,000,000

30,000,000

4,000,000 20,000,000 3,000,000 10,000,000

2,000,000

0

1,000,000 0

-10,000,000

Price (Mean) of Artist Google Rank(12/2/2011) Linear (Price (Mean) of Artist) Linear (Google Rank(12/2/2011))

31

CHAPTER TWO

WHY DO WE GO TO ART MUSEUMS? IS FREE ART MORE ATTRACTIVE?

Abstract This study shows how the personal attributes of state residents and different admission policies (free versus charge) affect museum attendance of state residents and visits at museums in a state. Using three analyses for both state residents’ and tourists’ art museum attendance, we found that as the percentage of total museums that are free in a state increases, the probability of going to art museums/galleries by people who live in that state increases. On the other hand, analyzing data with the tourist attendance leads us to conclude that free museums have a negative effect on the number of attendance. Our results also support the importance of location for the art museums. Having more than one art museum in close proximity increases art museum attendance.

JEL Classification: Z11, D11, D12, H41, L33

32

1. Introduction to Art Museums The United States is one of the most important countries in terms of art. There are hundreds of museums with the world famous painters and world famous paintings. The top art museums in the United States feature a remarkable collection, including the works of Picasso, Pollock, Van Gogh, Monet and etc. Many of the art museums’ structures are a piece of art in itself such as J. Paul Getty Center in Los Angeles and Guggenheim in New York. If the art museums are a vital part of the United State’s culture, do the residents take an advantage of this treasure? Does the museum behavior affect local participation? In general, museums can be considered as public good since some museums do not charge any admission fee. Therefore, no one is excluded from using it. This system applies no entrance fee and there is no donations box where the art participants make contributions. On the other hand, it is possible to reduce other persons’ use of museums by charging an admission fee but others are not completely excluded. However, the admission fee charged by museums may differ for different characteristic of participants as almost all museums today have a price discrimination policy for seniors and students. Some museums offer a suggested level or range of donations. Like museums that charge admission fee, the museums with suggested donations give tickets for the admission. This system is to insist on some payment, but to leave the amount to the visitors (O’Hagan, 1995). Moreover, there is another type of museums which accept donations. These museums neither charge any admission fee nor have any suggested donation amount for contributes. Decision of donating or not donating is left to contributors’ freewill. Museums can be categorized by three types. Public museums’ objective is number of attendance other than profit. They have little incentive to gain additional income and to minimize the cost (Frey, Metter, 2006). They get funds to cover the expenses. While the public museums

33

do not necessarily target to have profit, the private profit museums’ objective mostly depends on increasing their income. The objective of the museums which accept donations is to attract donors and to increase the amount of donations. Private nonprofit art museums’ income depends on admission charges, donations, the profit from gift shops and restaurants and funds from the government. Public museums in Washington D.C. are supported by the federal government and other public museums are supported by state and local governments (Feldstein, 1991). The support to arts also comes from The National Endowment for the Arts. NEA directs three basic types of financial support for the arts: direct public grants awarded by the National Endowment for the Arts and by state and local arts agencies. For example, The National Gallery of Art received an appropriation of $97 million in 2006 from the US Congress; arts and culture subsidies from federal agencies other than the NEA; and private donations.1 These sources themselves generate around 56 percent of total funding of U.S. nonprofit arts organizations. The rest of the revenue comes from the ticket sales and subscriptions.2 The discussion on the government support for the arts has been an issue in literature. Clotfelter (1991) addresses three arguments which justify the government intervention. First, art and art institutions produce beneficial externalities. Second argument emphasizes the responsibility of the government to protect cultural heritage for the future generation. Last justification for supporting the art is that government provides art museums for free because there are so important that everyone should have them regardless of ability to pay. Charging, not charging and setting entrance fee have been discussed by many authors. Since there exists positive externalities, we should consider the individuals in society and the museum should be reimbursed with tax money. Also, because of the low or zero marginal costs 1 2

“How the United States Funds the Art”, National Endowment for the Arts, 2007, http://www.nea.gov/pub/how.pdf “How the United States Funds the Art”, National Endowment for the Arts, 2007, http://www.nea.gov/pub/how.pdf

34

of participants, a zero price is efficient (Frey, Meier, 2006). However, free admission is not applicable to all art museums. Demand for museums is price elastic. Therefore, an increase in admission fee could reduce attendance. There are three important (groups of) players in the analysis of art museums/galleries. The first player is the visitors who are local, domestic or international tourists. The second player is art museums themselves, which can be public or private; museum or art gallery. The last player is the state/federal government or other parties, who provide funding and set policies. Other studies have mostly considered the museum structure, funding to museums and pricing or not pricing the admission. We add to these efforts emphasizing additional supply side characteristics of museums and specifically consider how they impact consumer behavior. This paper seeks to understand the players; how the personal attributes of state residents and how the characteristics of art museums and states influence the art attendance. Our focus is to analyze the impact of different types of art museums in terms of the admission fee policy to domestic public participation of state residents and also overall art museum visitors.

2. How do Art Museums and Art Galleries Work? Art museums can be nonprofit organizations or publicly owned profit organizations. They display art related objects but mostly paintings. According to Association of Art Museum Director (2007), more than 90% of art collections in art museums were donated by individuals. The rest of the art pieces were bought by museums. Therefore, donations actually make significant contribution to the success of the museums. What do/must art museums try to achieve? Stephen Weil (2002) considers this question and discuses that art museums must have social outcome to make a difference for the audience.

35

He states that art museums need to compete with each other not only for number of audience or exhibitions but also for the outcomes. He says “Outcomes are what happens outside an organization as a consequence of distributing the goods and/or services that the organization produces.” The social outcome is also one of the main reasons to support “not charging”. Weil defines the social outcomes as skills, knowledge, values and so on. Of course, they must provide beneficial outcomes to the public. According to Weil, if the museum is actually worth what it costs depends on the people. Therefore, it depends on audiences’ needs and their satisfaction. In this point of view, the satisfaction of visitors comes with the success of art museums. However, measuring the satisfaction and success is not easy. Museums survey their audiences to see their likes and dislikes. For example, Metropolitan Museum found that good special exhibitions increase audiences’ visiting frequency and lead them to spend their time in museum’s permanent collections (Luers, 1991). In reference the Weil’s paper, Anderson (2004) discusses measurement of success of art museums. He says that there are three primary (leading) indicators of success in art museums: the number of major shows (exhibitions), the number of visitors and the number of members. Most exhibitions are temporary and run for short length of time. The admission fees for special exhibitions are usually additional and hence generate unstable revenues for the museums. Anderson (2004) discusses that the existence of direct and indirect costs of having an exhibition in the museum. Direct costs are insuring, shipping, installing and spending in opening nights. Indirect costs are opportunity cost, marketing, part time staff and so on. According to Anderson, few exhibitions result in a surplus. In other words, it is more essential to focus on permanent collections other than temporary exhibitions. In our paper, we only consider the permanent collections for our analysis.

36

For the residents’ participation, the probability of visiting local art museums might be high if the museums in their local area have good quality service and if visitors are satisfied after each visit; however tourists would prefer the museums with high international reputation rather than experienced quality. The visitors’ satisfaction from the museums can be also measured by the average length of a museum visit. According to Anderson (2004), the total numbers of tourist visitors and the number of volunteers are measure of reputation and reputation is a measure of the success of art museums. In our study, we consider local visitor demographics with the demographics of the local population for more representative results. The one of the components of art museum finance is income. Public support mostly comes in the form of taxes (Schuster 1998). Nonprofit art museums are exempt from federal income tax. Tax policy also changes the incentives of donors and investors because of the deduction against income tax or estate tax (Fullerton 1991). Gifts can be considered just as another expenditure of the donors hence by donating to the museums they also gain benefits such satisfaction, membership and prestige. Fullerton discusses the effectiveness of the tax policy on art museums. He states that tax system provides incentives for charitable gifts. As donation is a significant factor for art museums, increasing in incentives would improve the quality of art museums. The tax money flow from public to art museums is shown as follow: Taxpayer’s Money Appropriations Committee

Government Agency Grant Opportunity Program

Spending Authority Funds Awarded to Museums

Peer Review Grant to Museums 37

Line Item Funds Awarded to Museums

Source: Institute of Museums and Library Services3

Museums can receive grants from one or more sources showed in this figure. Grants depend on museum type and level of government. In 2008, 62.5% of art museums reported that they received some government support. 4 72.2% of art museums reported state support. National Endowments of the Art has to allocate 40% of its grant funds to state and regions. The NEA reported that both federal and state government appropriations to the arts have declined since the 2007-2009 economic recession. The NEA’s annual appropriations were increased by $10.3 and $12.8 million in 2009 and 2010, respectively. However, they were reduced in 2011 by $12.8 million and in 2012 by $8.7 million. Annual appropriations to state arts agencies have continued to decline in recent years. In 2009, appropriations declined by $25 million, and were again reduced by $37 million in 2010, $17.9 million in 2011 and $13.9 million in 2012.5 Figure 1 shows the total appropriations to the National Endowments of Art and the State Agencies between 1970 and 2012. According to Fullerton (1991), in order to support government for subsidies to art museums, the cost of the service for art museums must be compared with the value of the service for the visitors. Also, art museums have some external benefits to the visitors or positional future local visitors. For example, national prestige and educational benefits to public support the government subsidy to art museums. Art museums revenue comes from audiences, government subsidies, cooperate sponsorship and endowments. Admissions, gift shops, bookstores and restaurant revenues are related with number of audience. Krens (1991) states that after having reduced government 3

Exhibiting Public Value: Government Funding for Museums in the United States December 2008 Exhibiting Public Value: Government Funding for Museums in the United States December 2008 5 National Endowments for the Arts (2012), “How the United States Funds the Art”, http://www.nea.gov/pub/how.pdf 4

38

support in the 1980s, museums started competing with each other to get funds from the pool. Declining government support to arts caused increase in cooperation supports. Thus, when the government support is not enough for museums to finance then art museums look for other sources from co operations. Banks support museums with a combination of permanent loans not only in the US but also in other countries. For example, in 2008, Deutsche Bank and the Städel Museum agreed that the bank would give the Städel Museum 60 paintings and sculptures, 161 originals on paper and 379 prints.6 By this way, they became partners and the museum has a gallery with the name of the bank. The income from endowments covers the significant amount of operation costs. However, the percentage of operating expenses covered by endowments was higher in the past than now (Krens, 1991). The other component of art museum finance is expenses. The costs of art museums include employment salaries, technology and maintenance expenses such as museum or exhibition decorations. Art museums insure the art works they exhibit. The insurance cost depends on the value of art work. As previously discussed, some art museums have admission fee policy while others have either voluntary donations or free admission policies. Most of art museums with admission fees have discrimination price policy. The categories with discounted price include children, students, seniors, local residents, groups, disabled and family groups. Steiner (1997) defines two separate goods in art museum consumption; weekends and weekdays. The time and the day for audience to visit the museums depend on opportunity cost, income elasticity and price inelasticity of audiences.

6

Deutsche Bank, Deutsche Bank Collection at the Städel Museum, https://www.db.com/csr/en/art_and_music/staedel.html

39

3. The Model Hansen (1997) presented the individual’s utility function of the Royal Theater in Denmark and this model can be modified for art museums. Similar to Hansen (1997), Frey and Meier (2006) defined social values created by museums, which are not measured in monetary terms. We developed resident’s utility function for free and charge art museums. Addition to the visitors’ value, non-user’s value is also included to residences’ utility function. Suppose that a consumer’s preferences can be represented by the utility function: U  U ( A  A ,  , F ( F , ), E, Z )

(3.1)

Where A  value from visiting the art museum with admission fee; enjoying the art objects and

the service in the museum. F  value from visiting the free art museum; enjoying the art objects and the service in

the museum. A, value from visiting the art museum with admission fee, and F value from visiting the free art museum include benefits such as:

As , Fs  social benefit; benefit from socializing Ae , Fe  educational benefit; learning art Ap , Fp  pleasure; benefit from seeing art objects E= value from existence of the art museum: This is a non-user benefit from knowing that an art museum exists in their local area. Even though the residents do not visit the museum in their state, they still can gain from the museums. E, value from existence of the art museum, includes benefits such as:

40

E pr  prestige; international or domestic recognition of the museum from domestic or foreign tourists.

Ec  cultural benefit; national benefit, sense of culture Eov  optimal value; the probability of going to the art museum in the future. Z  all other goods and services

  level of quality of the art museum   consumer’s taste parameter; related with the art exhibited in the museum such as

modern art or old master paintings. This utility function can be maximized with regard to the prices and the consumer’s income: max U  U ( A  A ,  , F ( F , ), E, Z )

(3.2)

The budget constraint is represented as:

M  PA A( AS , Ae , Ap , A , )  CF F ( FS , Fe , Fp , F , )  PE E( E pr , Ec , Eov )  Pz Z

(3.3)

where M is individual’s income, PA is admission fee, PE is taxes for public subsidies and PZ is price of other goods. There might be some cost of going to museum other than admission fee. Therefore, even for free museums, there occurs cost such as transportation cost and opportunity cost, CF . The consumer has an expenditure function which indicates the minimum amount of income to reach a certain level of utility. Consumers try to find the cheapest way to obtain the utility. The consumers’ willingness-to-pay problem can be presented such that the consumer has to minimize the expenditure function: e( PA , CF , PE , Pz ,  A0 ,  F0 ,U 0 )  min{ pA A( AS , Ae , Ap , A , )  cF F ( FS , Fe , Fp ,  F , )  pE E ( E pr , Ec , Eov )  pz Z | U ( A  A ,  , F ( F , ), E, Z )  U 0}

41

(3.4)

Not all households have the perfect information. If the households are not aware of the art museum in their states, no value from existence of the art museum (E) are plausible for that museum (Whitehead and Blomquist 1991). We define the cases when the consumers are aware of the museums and when they are not. This happens if the consumers visited the museum before or used any resources provides information about the museum. Then, the information about the museum, IE : IE  1

if E >0

IE  0

otherwise

(3.5)

where IE=1 indicates information about the museum and IE=0 indicates no information about the museum. Frey and Meier (2006) discussed the role of directors of art museums. They stated that the managers of public museums do not try to get additional income like other museums. The energy and the resources are limited and not allocated for this goal because they are not dependent on income from entrance fees or shops (Frey and Meier, 2006). As it is mentioned before, most museums try to achieve their goal of increasing their income. Therefore, they will make an effort to attract visitors. In this point of view, we can come to a conclusion of that the quality might depend on museums’ goals. From the incentives of directors, we can assume that

 A   F . We also assume that benefit from existence of any museums is same for private museums, private nonprofit museums and public museums. Policies impacting museums can change incentives and the quality of the museum. The use value can be defined as the willingness to pay to avoid a price increase that would cause not going to art museums. Assume that quality of museums with admission fee increases after the policy implication,  A1   A0 :

42

Total value includes use value and existence value.7 (Madariaga and McConnell 1987; Smith 1987) * 1 0 0 0 0 0 WTPChTotal arg e = e( PA , CF , PE , Pz ,  A ,  F ,U | I E  1)  e( PA , CF , PE , Pz ,  A ,  F , U | I E  1)

(3.6)

where PA* is the choke (reservation) price which is the highest price makes the charge art museum demand zero and WTPChTotal arg e is the change in expenditure necessary to stay at the same utility level after the development of charge art museum. Now assume that  F1   F0 Total WTPFree = e( PA , CF* , PE , Pz , A0 , F1 ,U 0 | I E  1)  e( PA , CF , PE , Pz ,  A0 , F0 ,U 0 | I E  1)

(3.7)

where CF* is the choke (reservation) price which is the highest cost (transportation, etc.) makes Total the free art museum demand zero and WTPFree is the change in expenditure necessary to stay at

the same utility level after the development of free art museum. Total To compare the total values of free art museums and charge art museums, WTPFree , and

WTPChTotal arg e , we should compare the expenditure functions for these two types of museums. We assume that  A1   A0 and  A   F , then,

e( PA* , CF , PE , Pz , A1 , F0 ,U 0 | I E  1)  e( PA , CF* , PE , Pz , A0 , F1 ,U 0 | I E  1) and Total WTPChTotal arg e > WTPFree

(3.8)

We conclude that the change in expenditure necessary to stay at the same utility level is greater for charge art museums than for free art museums. In other words, the willingness to pay for the charge museums is greater than the willingness to pay for the free art museums in the case of  A1   A0 ,  F1   F0 and  A   F . The quality of charge museums affects the public attendance. If

7

See Appendix

43

charge museums in the state increase the quality more than the free museums, this will increase the willingness to pay of state residents for charge museums more than free museums. Despite the clear importance of quality in art markets, our empirical model does not include an explicit quality variable due to lack of information. The model will show residents’ participation to art museum and the art museum success without the assumption that  A1   A0 ,

 F1   F0 and  A   F . The empirical model is developed to see the residents demand to art museums and how they are sensitive to free and charge art museums.

4. Empirical Model The first part of our study, probit model is used for individual respondent behavior. The question is how the characteristics of residents affect the probability of their museum attendance. We also consider the museum admission policies with the percentage of total museums that are free. As we did in the theory part, we compare the free museums and charge museums for public attendance. The analysis is based on the cumulative normal probability distribution. The probit model is used to model binary outcome variables. The dependent variable is either “Yes” or “No” answer to a survey question which was asked to the residents of a state in 2002 and 2008. The binary variable takes on the value of zero or one. Probit modeling can be formulated: K

yi*  0    k X ki   i

( 4.1)

k 1

X ki  K independent variables for individual i

0  intercept k  parameter for the effect of x on y*

 i  error term for individual i 44

1 yi   0

if y*  0 otherwise

(4.2)

We could define the binary variable:

1 yi   0

if i th person went to an art museum or gallery in last 12 months otherwise

The probability of going to museum can be shown as:

prob( y  1)   ( yi* )

(4.3)

prob( y  0)  1   ( yi* ) Where,  ( yi* )  Cumulative normal distribution function The number of art attendance from public may vary from individual to individual. Many reasons can affect the attendance habit. In this point of view, we use ordered probit model for the frequency of going to art museums/galleries. The model is specified as: K

Yi*  0    k X ki   i

(4.4)

k 1

X ki is k independent variables for individual i. Yi * is an unobservable latent variable but what is observable is Yi which is the frequency that the individual visited art museums/galleries with the ordered categories: None, 1-3 times, 4-10 times and more than 10 times. These ordinal categories were not estimated from the data, but correspond to categories used in prior. These ordinal categories were not estimated from the data, but correspond to some of the categories used in prior studies such as Falk and Falk (2011).

0 1  Yi   2  3

if

Yi *   1

if

 1  Yi *   2  2  Yi *   3  3  Yi *

if if

(4.5)

45

and prob(Y  " No times" or 0)   ( 1   ' x ) prob(Y  "1  3 times" or 1)= ( 2   ' x )- ( 1   ' x ) prob(Y  " 4  10 times" or 2)= ( 3   ' x )- ( 2   ' x )

(4.6)

prob(Y  " More than 10 times" or 3)=1- ( 3   ' x )

 1 ,  1 ,  1 ,  1 are the unknown parameters to estimate and  is the cumulative function of a normal distribution. Since the estimated  value in both probit and ordered models does not estimate the change in probability of going to art museums due to the unit change in the independent variable, we use marginal effects to explain the model results. For example, the marginal effect of changes in x j on probability of going to museum 4- 10 times is:

Prob(Y  2)  ( ( 3   ' x)- ( 2   ' x))  j x j

(4.7)

In the second part of our study which focuses on attendance at a more aggregated level than in the first part, we use random effect model to estimate the variables that are constant within unit (time invariant variables). The information is collected from art museums/galleries. The dependent variable is the number of attendance to art museums/galleries in 2002 and 2008. We use the logarithmic attendance as a dependent variable. The model is as follow: L

Ln "Number of Attendance to an Art Museum"mt  0 + l Almt  um  emt

(4.8)

l 1

where m is museum and t is time. A includes explanatory variables at the level of characteristics of a museum. u is an individual effect and e is an idiosyncratic error term. Finally, the last part of our analysis includes information for each state in the U.S. The independent variable is the average of total attendance to art museums in the state. OLS is used to estimate the model to see the time effect. The model is as follow:

46

R

Ln "Average of Total Attendance to Art Museums in a State" j  0 +  r Crj  v jt r 1

where j is state and v is the error term . C stands for explanatory variables at the level of characteristics of a state.

5. Data In this paper, we have three different analyses to see public effect on art museums attendance and general art museums attendance. First data is collected from individuals. Second data is collected from museums. Finally, third data is for states. 5.1 State Residents’ (Public) Participation to Art Museums Household data obtain from national population survey are analyzed to reveal important factors influencing consumer decision to go to art museums and galleries. The data is called “the Public Participation in the Arts Supplements” and conducted by the U.S Bureau of the Census. In the survey (Current Population Survey), around 60,000 households were interviewed in May in 2008 and 57,000 households in August in 2002. The questions for art participation were asked of these Current Population Survey respondents. The households for the survey were randomly selected household member aged 18 or older from about one-fourth the sampled CPS households. The demographics included region of residence, employment status, disability status, marital status, household size, the presence of children in the household, education, age, gender, income and race. We created a state “Days of Snow”8 variable in the data to see the role of the location in attendance. We define the “Days of Snow” as average number of days it snows. Oster and

8

http://www.currentresults.com/Weather/US/average-snowfall-by-state.php

47

Goetzmann (2003) used the mean January temperatures for the tourist demand. However, this does not necessarily show the tourism effect on the attendance. It is mostly about how the location develops and gets adjusted due to the weather conditions. We suggest looking at and comparing the museums with cold winters and the museums in warmer places in terms of collections and employments. An alternative measure of importance of location and tourism is the ocean coast. We look at the places which have coast to the ocean and see how this affects the public participation to the art museums. In addition to the demographic variables, we also consider tax revenue per capita by state for all models. Since the money from tax payers is one of sources for museums, we can think this money as indirect public payment (or maybe sunk cost) to art museums. We use tax capita by state to see how residents use this indirect payment and benefit from art museums. The variable of particular interest in this paper is “percentage of free art museums” in states. We try to answer how the free museums rate affects the public participation. The percentage of free art museums is the proportion of free art museums over total art museums in the state. However, the CPS data does not distinguish the visits whether they happened during domestic or international travels or they happened to the local art museums or galleries. To deal with this issue, we use “number of neighbor states”. We want to see if the more neighbors the state has, the less probability the residents of the state go to museums. Hendon (1990) analyzed the survey data from 1982 to understand the art museum visitor and distinguish the visitor from those who do not go to art museums. In addition to Hendon, we also try to understand how percentage of art museums, which established for different goals, affects the behavior of residents of the states. The variables do not allow us to pool the data

48

because of the differences between variables depended on time. We run two different regressions for 2002 and 2008. 5.2 Art Museums Data The data for art museums was collected from “Official Museum Directory 2002” and “Official Museum Directory 2008”.9 The 2002 directory contains listing over 8,100 museums and the 2008 directory contains more than 10,500 in many different fields. Most of the information is reported by the museums including museums’ addresses, founded year, number of employees, facilities, attendance numbers (visitors) and type of the museum such as historical, art or zoo. Open hours and admission price are also listed in the directory. The panel data includes 1890 observation from 2002 and 2008. If the museum is free, it says “No charge”. If it is free but accepts donations, it says “No charge; donations accepted”. If the museum charges an admission free, then they announce the price and only few museums announce suggested donation amount. In second part of our study, the focus is on what the affects attendance at art museum. The data includes attendance of local, domestic and international visitors. Similar to Oster and Goetzmann (2003), we created a dummy for university-based museums and galleries. Additionally, we categorize the museums as public and private organizations. We also create a dummy variable to capture the effect of free museums on attendance. We use the admission fees for the non-free museums. Because of the multicollinearity problem, we have two different models with “free” dummy variable and with “admission fee” variable. The variable for admission fee includes ticket prices for charge museums and zero for free museums.

9

Published by American Association of Museums and National Register Publishing

49

The same problem we have with the first data (CPS) also occurs here (Official Museum Directory) for local and tourist attendance. The data does not include the visitor’s location information. To deal with this problem, we use two different distance variables. The first variable shows how far the museum is to the largest (population wise) city. The second variable shows how far the museum is to the second largest city. The variable is created by using the zip codes of museums and the zip codes of the downtown of the largest cities in the state. We also use zip codes to count how many museums are located in the same area and created a variable which indicates the number of museums which are close to each other. Thus, our purpose is to see if it is convenient to have more than one museum in the same location and how this affects the attendance. “Days of Snow” variable is used to measure how museums do if they are located in cold states. We created “Art Lover” variable which is the percentage of people who participated to the survey and went to art museums in last twelve months over all people who answered the question. The “Art Lovers” variable is used to see if the characteristics of people who live in the state where the museum locates have any effect on overall museum attendance. It shows the relationship between the art lover state and the attendance to the art museums. We use the “Art Lover” to see If the percentage of people who go to art museums increases in the state where the museums is located and how this affects to overall art museum participation for the museum. This variable is to see if there is any local effect on attendance. 5.3 States Data The third data set we used in this paper is collected from states’ information. The data was mostly generated from the first part (individuals) and the second part (museums). The dependent variable, total number of art museum attendance, is from the second part of the 50

analysis. We use the “Art Lover” variable from the first part. The purpose of this analysis is to see how local participation affects the art museums attendance in states. By using the state data, we clearly see the tourism effect on attendance. Number of attendance to national parks 10 and dummy variable for the most visited states11 are used to capture the tourism effect. We also introduce a dummy variable for 2008.

6. Results The results are developed in three parts with three different analyses. The first part is in household, the second part in art museum and the third part in state level. 6.1 State Residents’ (Public) Participation to Art Museums The data used in this part is from U.S. Census survey named “the Public Participation in the Arts Supplements”. The regression includes demographics variables and interaction terms. Table 5 reports the probit estimation coefficients and shows how the individual’s characteristics, individual’s location and the characteristics of individual’s state affect public participation to art museums. The probability of going to art museums increases with income, age and higher education. Table 5 shows that being a woman, being a white, living in metropolitan area and being close to water each has positive effect on the probability of going to art museums. On the other hand, disability, number of household, being Hispanic, working fulltime and working in private sector each has negative effect on probability of participation the art supplements.

10

Source: National Park Service, U.S. Department of the Interior, https://irma.nps.gov/Stats/SSRSReports/System%20Wide%20Reports/YTD%20Report%20%28By%20Park-State-Park%20TypeRegion%29 11 Source: U.S. Department of Commerce, International Trade Administration, Office of Travel and Tourism Industries, and the Bureau of Economic Analysis (BEA), May 2010, .

51

Being married without children and having a child but not married decreases the probability of art participation. On the other hand, the interaction term shows that married people with a child also decreases the probability for art attendance even more. Also being employed (not in principal city) has negative effect on art attendance while being employed in principal city increase the probability of going to art museums. Age square shows that aging increases the probability at a diminishing rate. Another interaction term, education square, measures the effect of higher education on the probability for art attendance. It shows that people who have some college education but not degree or higher education increases the probability. Also, people who live in states which have more than four neighboring states, most likely go to art museums. The main question we explore using the resident’s survey data concerns the impact of free art museums located in the resident’s state: Does the percentage of free art museums affect the visitor decision? We see from the Table 5 that the probability of going to art museums increases with the increase in free museums rate in the state. If the percentage of museums that are free is higher in a state, then respondents from that state are more likely to have gone to museum (of any kind) during the survey year. We see this positive effect in 2008 data. However, we see that the effect of free museums in 2002 is negative. We can conclude that the economic conditions in 2008 have increased the use of free museums. The results further suggest that the tax revenue per capita matters in terms of public participation. People who live in a state which has high tax revenue per capita were more likely to go to art museums in 2008. Again, this effect is negative in 2002. Ordered probit model is used to analyze public participation behavior. The model examines the effects of explanatory variables on participation in art museums similar to the prior model. However this model uses information on the number of times in 2008 the respondents

52

went to the art museums, not just whether they went or not. Based on the answers of the question, “How many times did you go to art museums/ galleries in 12 months?” there are four ordered categories for the dependent variable: None, 1-3 times, 4-10 times and more than 10 times. To better interpret the parameter estimates for the ordered probit estimation, Table 2 provides the marginal effect of the ordered probit for public art museums attendance in 2008. 13

12

The marginal effects are presented for each level of the dependent variable, as the marginal

effects differ (in direction and magnitude of impact) by level. Table 6 shows that the marginal effects of income, education, living on coast, the percentage of free museums and living in metropolitan is positive. We find that the probability of art museums/galleries attendance increases with household income, education, living on coast, the percentage of free museums and living in metropolitan in the state for all categories besides the category “None”. For example, an increase in household income by $1000 (from 50,180 to 51,180) will decrease the probability of non-attendance by 0.135 percentage points. An increase in household income by $1000 increases the probability of art museum attendance of “1-3 times” by 0.108 percentage points, the probability of art museum attendance of “4-10 times” by 0.024 percentage points and finally the probability of art museum attendance of ‘10 or more times” by 0.003 percentage points. The marginal effects for gender show that being a female reduces the probability of non attendance by 3.7 percentage points, whereas the probability of the frequency of attendance for the three categories increase by 2.95, 0.66, zero percentage points. Moreover, being a disable person increases the probability of never attendance to art museums by 8.1 percentage points.

12 13

See Appendix for the marginal effects from ordered probit model for 2002 See Appendix for the ordered probit results for 2002 and 2008

53

Furthermore, an increase in the percentage of free museums in the state by 10 percent (from 52.5 percent to 58 percent) will decrease the probability of non-attendance of residents from that state by 14.9 percentage points. Also, this increases the probability of art museums attendance of “1-3 times” by 11.8 percent, attendance of “4-10 times” by 2.67 percentage points and “10 or more times”14 by 0.37 percentage points. As we see from the results, the marginal impact of an increase in the percentage of free museums in the state on “None” category is quite large. The marginal effects for variables that are nonlinear in the model (interaction terms, squared terms, etc) are different than the numbers on Table 6. The marginal effect of being an employee in a principal city for the probability of non attendance is -0.0274.15 In other words, being an employee in a principal city reduces the probability of non attendance by 2.7 percentage points. For 1-3 times, 4-10 times and more than 10 times visiting, it increases by 2.1, 0.5 and 0.08 respectively. While being married without children increases the probability of non attendance by 4 percentage points, being single with children also increases the probability by 4 percentage points. However, there is no statistically significant effect of being married with children on the probability of non attendance. Age was included in the model with a linear and squared component, to allow for a nonlinear impact on the probability of attendance. Evaluated at the sample mean age of 47.7 years, the calculated marginal effect for non attendance is 0.00192.16 An increase in age by 10 percent above the mean (from 47.7 to 52.5) will increase the probability of non-attendance by

14

The data is restricted with 25 times art museum visiting. 15 See Green and Hensher (2009), “Modeling Ordered Choice”, Page 115-116 for the calculation. 16 The calculation is as follow; (Age)=-0.0019+2(47.7)(0.00004)

54

0.192 percentage points. However, there is no age impact on the probability of non-attendance or other categories for younger residents in 2008. Age has positive effect on attendance in 2002. The impact of neighboring states on probability of art attendance is measured for minimum and the maximum number of neighboring states. For example, living in a state which has one neighbor state increases the probability of non-attendance by 2.5 percentage points. Living in a state which has seven neighbor states decreases the probability of non-attendance by 1.1 percentage points. The analysis was completed for the 2002 data and is reported in the appendix. Although we found the positive effect of an increase in the percentage of free museums in 2008, there is an opposite effect in 2002. An increase in the percentage of free museums in 2002 by 10 percent increases the probability of non attendance by 6.9. For the remaining three categories “1-3 times”, “4-10 times”, “10 or more times”, the decreases in probability are 5.3, 1.4 and 0.26 percentage points respectively. 6.2 Art Museums’ Characteristics Effect on Art Participation Table 7 shows the results from the random effect model for the museum level with the dependent variable of ln(number of visitors to the art museum). The number of visitors includes all attendance to an art museum in 2002 and 2008. Art museums are categorized by their characteristics (age, free/charge, museum/gallery, number of fulltime employees, etc), their location and the state where the museum locates. There are two models in this part. Model 1 in Table 7 is with free art museum/gallery dummy variable. The admission fee (ticket prices for adults) was used in Model 2 instead of the free art museum/gallery dummy variable. Age of an art museum has positive effect on the number of attendance. The older the art museum is the more people visit. Art galleries have negative effect on the number of attendance. 55

If the organization is an art gallery, this decreases the attendance. Art museums take more visitors than art galleries. Also, Similar to Oster and Goetzmann (2003), museums at university and college affect the attendance in negative way. They discuss that college museums may focus on education and connoisseurship. It seems from the results that public museums do better than private museums in terms of the number of audiences. The number of full time employees at an art museum shows how big the art museum is. The effect of the number of full time employees is small but significant. Museums which have more full time employees have more visitors. In contrast to the results in part 1 where free museums (%) had a positive impact on the probability of attendance, free museums have a negative effect on attendance in Model 1. Free museums deliver the smaller audiences. This might be due to the tourism effect. The data does not include any quality variables for art museums due to the lack of information. We assume that the effect of free museums might be a signal of poorer quality. However, there might be other quality variables related with museum collections such as artists and number of paintings that are not included. Hence we can only speculate how certain results are linked to quality. Local people might not consider the quality for the free museums. The cost (opportunity cost) of going to free museums is lower for local people than for domestic or international visitors. Similar to Model 1, Model 2 shows that an increase in admission fee increases the art museums attendance (measured as ln of the number of visitors). The square of admission fee is significant and negative. Ticket prices increase the percentage of number of attendance with diminishing rate after 12.5 dollars. 17 The distance of an art museum to the biggest city and to the second biggest city also affects the number of attendance. The farter a museum to the big city is the less attendance the 17

2

This is the attendance maximizing price which can be charged. aP+bP . By maximizing, a+2bP=0 or –a/2b

56

museum has. However, the number of art museums at the same location with an art museum increases the attendance to that museum. It may be convenient for the visitors to visit many museums at the same day especially if the art museums are close to each other. It shows the importance of location for an art museum. It is better to establish an art museum where there are one or more art museums. The number of days it snows increases the number of attendance to an art museum. If the museums locates in the area which has long winter time affect the museum’s attendance. This finding is similar to the first part with local participation. This may show either museum in cold states develops more and focus on art more than warmer places or the people who live or visit the cold states prefer to go to art museums. Although the probability of going to art museums increases for the local people who live close to ocean or big lake, coast does not affect the art museums’ visitor attendance. Art museums which are located in those states do not statistically get affected by being close to coast. We use the “Art Lover” variable which represents a percentage of respondents who went to art museums in 2002 and 2008. This variable was created from the survey data for residents. We found that there is statistically significant effect of residents to the museums attendance. “Art Lover” variable shows the percentage of going to art museums and the characteristic of people towards art. If the percentage of people who go to art museums increase in the state, this increase the percentage of number of attendance to the art museum which locates in that state. Also, the population of the state which the museum locates increases the number of attendance. There is no statistically significant effect of state size to the museum attendance.

57

6.3 State Characteristics Effect on Art Participation The third part of our analysis is aim to show the state effect on art attendance and to assess whether there were changes in factors impacting attendance over time. The dependent variable is ln(average art museum visitors in the state). The data for 2002 and 2008 are combined in one model to see the time effect. OLS results are presented in Table 8. This model includes variables that are categorized as State Variables, Location Variables and Time. The results indicate that there is statistically significant effect of the state on art museums attendance. Similar to part 2, we used the “Art Lovers” variable in part 3. The effect of residents on art attendance is positive. If the percentage of art lovers in a state is high relative to other states, the overall number of visitors tends to be higher, all other factors held constant. The result may support the local participation to local art museums. “Art Lovers” variable also shows how big the local support to art museum is. We see that the effect is very small. Thus, art attendance is mostly related with international and/or domestic tourism. Another variable which can be related with local participation is the percentage of free museums in a state. Converse to the first part where free museums (%) had a positive impact on the probability of attendance, we found that when we analyze the data that includes with all visitors, and include variables such as state population, there is no free museum(%) effect to the average art attendance at the state level. This might be due to international and/or domestic tourism effect. We showed in the second part in Model 2 that admission fee increases the audiences in art museums. Similar to this finding, we found that average price of the art museums increases the average art attendance in states. One of the tourist-related variables is the number of national park visitors. We show that the number of national park visitors has positive effect on the average art museum/ gallery attendance in the state.

58

Additionally, we use the variables “population” and “geographic size” to control how big the state is. The results show that size of the state does not have any statistically significant effect on art attendance, however the larger the population of a state, the greater is the total attendance at art museums in that state. As we mentioned before, public support for art museums and galleries come mainly in the form of taxes. We use tax revenues per capita in the states as a variable measuring public support. The results show no significant statistical relationship between tax revenue and art participations. Tax revenue variable includes revenues for other than art, and hence its lack of significance may be an error-in-variable problem and not an unimportant factor. In contrast, the state real income per capita shows a positive relationship with art attendance. The richer the state is, the more art visitors it has. One can think that the richer states invest more on art and have better museums so that they can attract visitors or people in richer states go to art museum more. Art participation also depends on the geographic/location characteristics of a state. The location variables are “number of neighbor states”, “regions”, “days of snow” and “most visited states”. The study shows that location is an important component for the attendance. If the state is surrounded by other states, this increases the art attendance in that state. In other words, number of neighbor states attracts the people to visit art museums. This might explain traveling between states and participation to arts. Moreover, days of snow has positive effect on art. If the state is located in a snowy area, this increases the attendance in that state. This result is consistent with the public in the first part and art museums in the second part. It seems that cold states focus on art to attract tourists and give their effort to improve the art participation. People who live or visit these cold states prefer to go to art museums/galleries.

59

Another tourist-related variable suggests that the most popular states have more art visitors. Region of location of the state matters, with states in the Midwest, West and South all having higher attendance than Coastal States, all else constant. In this part of our study, the data allows us to pool two years. The Chow Test was used to see if the two datasets from 2002 and 2008 are equal. The model indicates there are not structural changes over time so that we could combine the data, with a year dummy. By doing this, we see the time effect on art. Table 8 shows that art participation in 2008 was lower than 2002. We see the effect of recession in 2008 to art participation. Graph 1 exhibits that the total appropriations to the National Endowments of Art and the State Agencies were lower in 2008 than in 2002. The appropriations to the art were at its peak in 2002. It started to fall dramatically after then. In 2008, it was at the same level when it was in 1998. Our finding also shows that visitors’ supports to art museums decreased from 2002 to 2008 as well as government support to the art. We see that art is affected by the economic satiation of the country in both visitors and government sides. People do not go to art museums during recession as much as they go in other times. 6.4 Price Elasticity of Demand for Art Museums The price elasticity of demand for art museums shows how demand is sensitive to changes in own-price. In this part of our study, we exclude the free museums from the data to find the price elasticity for charge museums. Figure 1 and Figure 2 display the public and total attendance to art museums in selected states and all regions in the USA. The results show that the art museums are Giffen goods from the visitors’ point of view. In other words, when we include all the visitors with tourism and public attendance18, we see that

18

Public attendance was measured with the percentage of people who went to art museums in last 12 months. We assume that people who said “YES” to the survey question went to a charge museum at least once.

60

price elasticity is positive. When the price increases, the demand for charge art museums also increases. However, Figure 2 shows that demand for art museums is inelastic in general. The change in price does not affect the demand much. Moreover, the price elasticity of public (local) demand for the art museums is negative. It shows that the art museums are ordinary goods. Although it is relatively more elastic than general demand, it is still not very sensitive to change in price. Among the four regions, a price change for art museums is more sensitive to people who live in south and it is elastic.

7. Conclusion This study investigates the attendance at art museums/galleries. We particularly focus on public attendance and free museums. Our work includes both state residents’ and tourists’ art museum attendance. We analyze the attendance with museums and individuals’ information. Our theoretical model suggests that if charge museums in the state increase the quality more than the free museums, the willingness to pay of state residents for charge museums increases more than for free museums. The empirical part showed that free museums have an impact on local art participation. If the percentage of free museums in a state increases, the probability of going to art museums/galleries by people who live in that state increases. This might explain how free museums can influence the public attendance. Also, the attendance to art museums increases with the percentage of public participation to art museums in that states. This is also an evidence for the role of public participation to local (state) art museums and galleries. When we analyze the total attendance with local visitors, domestic and international tourists, we see that free museums actually have negative effect on attendance. Indeed,

61

attendance increases with admission fee. This result shows how big the tourism effect on art museums is. Moreover, there is no relationship between overall art attendance to art museums and the percentage of free art museums in the state. We conclude that while the free art is good for local participation, it might be bad for tourist perspective. Price of art can be an indicator of quality for the tourists who are interested in visiting art museums. We see the tourism effect on our third analyses with the states. The most visited states by overseas travelers increase the attendance to art museums. Two of the most interesting findings in our study are about location. The results support that the closer the museum is to the biggest city’s downtown, the greater the audience the museum has. Also, our study shows that having more than one museum in the same location increases the attendance. One can think that it is good to start an art museum or gallery in locations where there are more art museums around. People prefer to go to art museums if there are more museums to go in the same area. Also, there are correlations between the total appropriations to the National Endowments of Art and the State Agencies and the number of art attendance. In our study, we found that attendance was lower at art museums in the year of 2008, all else constant. The information we got from the report of the National Endowments of the Art shows that the total appropriations to the State Agencies was less in 2008 than in 2002. It means that 2008 was not a good year for supporting the art in the USA or for attendance.

62

Bibliography Anderson, Maxwell L. (2004), “Metrics of Success in Art Museums”, The Getty Leadership Institute, http://www.cgu.edu/pdffiles/gli/metrics.pdf Association of Art Museum Directors (2007), “Art Museums, Private Collectors, and the Public Benefit”, http://www.aamd.org/papers/documents/PrivateCollectors3.pdf Clotfelter, Charles T. (1991) “Government Policy toward Art Museums in the United States”. In M. Feldstein, ed., The Economics of Art Museums, Chapter 9, 231-270, University of Chicago Press, Chicago. Falk, Martin and Rahel Falk (2011) “An Ordered Probit Model of Live Performance Attendance for 24 EU Countries”, Austrian Institute of Economic Research (WIFO), First draft: March 2011, http://martin.falk.wifo.ac.at/fileadmin/homepage_falk/files/Falk_Falk_2011

Feldstein, Martin (ed.) (1991), “The Economics of Art Museums”, University of Chicago Press, Chicago Frey, Bruno S. and Stephan Meier (2006), “The Economics of Museums”, in Ginsburgh, V.A. and Throsby, D. (eds) Handbook of the Economics of Art and Culture, , Amsterdam: Elsevier, Vol. 1, 1017-47. Frey, Bruno S. and Lasse Steiner (2010) “Pay as you Go: A New Proposal for Museum Pricing”, CESifo Working Paper, No. 3045. Fullerton, Don (1991), “Tax Policy Toward Art Museums”, The Economics of Art Museums, http://www.nber.org/chapters/c11643

Hansen, Bille (1997), “The Willingness to Pay for the Royal Theatre in Copenhagen as a Public Good” Journal of Cultural Economics, Vol. 21, 1-28. Hendon, William (1990), “The General Public Participation in Art Museums: Visitors Differ from Non-Visitors, but Not as Markedly as Case Studies Have Indicated”, American Journal of Economics and Sociology, Vol. 49, No.4, 439-457. Krens, Thomas (1991), “The Economics of Art Museums”, University of Chicago Press, Chicago Luers, William H. (1991), “The Economics of Art Museums”, University of Chicago Press, Chicago Madariaga, Bruce and Kenneth E. McConell (1987), “Exploring Existence Value”, Water Resources Research, Vol. 23, 936–942.

63

Manjarrez, Carlos, Carole Rosenstein, Cleste Colgan and Erica Pastore (2008). “Exhibiting Public Value: Museum Public Finance In the United States” (IMLS-2008-RES-02). Institute of Museum and Library Services, Washington, DC http://www.imls.gov/assets/1/AssetManager/MuseumPublicFinance.pdf

National Endowments for the Arts (2012), “How the United States Funds the Art”, http://www.nea.gov/pub/how.pdf

O’Hagan, John W. (1995), “National Museums: To Charge or Not to Charge?”, Journal of Cultural Economics, Vol. 19, 33-47. Oste, Sharon and William N. Goetzmann (2003). "Does Governance Matter? The Case of Art Museums," NBER Chapters, in: The Governance of Not-for-Profit Organizations, 71-100 National Bureau of Economic Research, Inc. Schmalen, Richard (1972), “Option Demand and Consumer's Surplus: Valuing Price Changes Under Uncertainty”, The American Economic Review, Vol. 62, No. 5, 813-824. Schuster, Mark J. (1998), “Neither Public nor Private: The Hybridization of Museums.” Journal of Cultural Economics, Vol. 22, 127–150. Smith, Kerry V. (1987), “Nonuse Values in Benefit Cost Analysis”, Southern Economic Journal, Vol. 54, No. 1, 19-26. Steiner, Faye (1997), “Optimal Pricing of Museum Admission”, Journal of Cultural Economics, Vol. 21, 307-333. The Official Museum Directory (2002), 32nd ed. New Providence, NJ: National Register Publishing, 2002. The Official Museum Directory (2008), 38th ed. New Providence, NJ: National Register Publishing, 2008. United States Department of Commerce. Bureau of Census, United States Department of Labor. Bureau of Labor Statistics, and National Endowment for the Arts. Current Population Survey, May 2008: Public Participation in the Arts Supplement. ICPSR29641-v1. Ann Arbor, MI: Interuniversity Consortium for Political and Social Research [distributor], 2011-04-15. doi:10.3886/ICPSR29641.v1 United States Department of Commerce. Bureau of the Census, United States Department of Labor. Bureau of Labor Statistics, and National Endowment for the Arts. Current Population Survey, August 2002: Public Participation in the Arts Supplement. ICPSR03971-v2. Ann Arbor, MI: Inter-university Consortium for Political and Social Research [distributor], 2011-04-22. doi:10.3886/ICPSR03971.v2

64

Weil, Stephen E. (2002), “Are You Really Worth What You Cost, Or Just Merely Worthwhile? And Who Gets To Say?”, The Getty Leadership Institute, http://www.cgu.edu/pdffiles/gli/weil.pdf Whitehead, John C. and Glenn C. Blomquist (1991) “The Link between Behavior, Information and Existence Value”, Leisure Sciences, Vol. 13, 97-109.

65

Table 1- Descriptive Statistics for Household Data Variable

Obs

Mean

Std. Dev.

Min

Max

Household Income

30257

50,192.10

28,330.79

2500

87500

Age

33706

47.72

17.22

18

85

Number of Households

33706

2.76

1.46

1

15

Education

33706

40.10

2.74

31

46

Child

33706

0.30

0.46

0

1

Employed

33706

0.64

0.48

0

1

Disabled

33706

0.05

0.22

0

1

White

33706

0.91

0.28

0

1

Married

33706

0.59

0.49

0

1

Female

33706

0.54

0.50

0

1

Principal City

33706

0.21

0.40

0

1

Hispanic

33706

0.09

0.29

0

1

Table 2-Frequency of Art Participation for State Residents Residents’ Art Participations

Freq.

Percent

Cum.

None

25,307

75.22

75.22

1-3 times

6,639

19.73

94.95

4-10 times

1,419

4.22

99.17

11-25 times

280

0.83

100

Total

33,645

Table 3-Descriptive Statistics for Art Museum Data Variable

Obs.

Mean

Std. Dev.

Min

Max

Number of Visitors

1,754

76,910.08

237,797

200

5,400,000

Free Museums

1,753

0.64

0.48

0

1

Adult Ticket Price

1,753

2.08

3.31

0

20

Fulltime Employees

1,715

22.54

82.31

0

1,783

University/College

1,754

0.31

0.46

0

1

Museum/Gallery

1,754

0.47

0.50

0

1

Age of Museum

1,754

50.84

33.44

2

236

Distance (Mile) to the Biggest City

1,754

105.10

104.62

0.1

674.41

Distance (Mile) to the Second Biggest City Number of Museums Around (with same zip code)

1,754

140.87

116.61

0.1

539.4

1,754

0.28

0.64

0

4

66

Table 4-Descriptive Statistics for State Data Variable

Obs

Mean

Std. Dev.

Min

Max

Total number of Museums

98

37.78

37.88

3.00

226.00

Average Attendance

98

60,970

27,522

13,930

152,238

Art Lovers

98

25.39

9.33

5.93

82.42

Coast

98

0.43

0.50

0.00

1.00

Average Ticket Price

98

2.05

1.23

0.21

6.00

Free Museums (%)

98

0.21

0.11

0.00

0.57

Days of Snow

98

18.88

16.51

0.00

66.40

Population

98

6,019,751

6,561,897

498,703

36,800,000

State Size

98

60,533.27

46,755.06

1,045

261,914

Number of Neighbors

98

4.29

1.52

1.00

7.00

Real Tax Revenue)

98

2,310

2,849

189

15,920

National Park Visitors

96

5,105,158

6,661,650

16,113

34,400,000

Most Visited States

98

0.27

0.44

0.00

1.00

Real Income

98

51,345

7,597

35,607

69,037

Table 5 - Probit Model for Art Museum Attendance in 2008 and 2002 2008 Did you go to an art museum or gallery in last 12 months? Yes

Standard Errors

Coefficients Individual’s characteristics Ln real income Number of Household Age Age*Age Female Disabled Education Education* Education Married Child Married*Child White Hispanic Employee Principal City*Employee Private Sector Employee Individual’s Location Principal city Metropolitan West South Midwest Days of Snow

2002

Coefficients

Standard Errors

0.2595*** -0.0529*** 0.0079* -0.0001*** 0.1429*** -0.3646*** -0.6832*** 0.0105*** -0.1440*** -0.1742*** 0.1286* 0.3476*** -0.0948** -0.1171*** 0.1552** -0.0803***

0.0213 0.0116 0.0045 0.0000 0.0241 0.0760 0.1081 0.0013 0.0323 0.0615 0.0669 0.0526 0.0464 0.0371 0.0626 0.0284

0.1890*** -0.0402*** 0.0114** -0.0002*** 0.1824*** -0.4199*** -0.3951*** 0.0070*** -0.0352 -0.1064* -0.0519 0.3171*** -0.1217** 0.0548 -0.0830 -0.1320***

0.0196 0.0114 0.0047 0.0000 0.0248 0.0827 0.1078 0.0013 0.0328 0.0563 0.0626 0.0506 0.0505 0.0376 0.0624 0.0294

0.0498 0.1071*** 0.2905*** -0.0489 0.0693 0.0025**

0.0538 0.0317 0.0438 0.0500 0.0469 0.0010

0.2010*** 0.1259*** 0.2509*** -0.0634 0.1067 0.0021**

0.0527 0.0314 0.0445 0.0511 0.0472 0.0010

67

Coast State Variables Percentage of Free Museums in the State Ln Tax Revenue Per Capita Ln Income Level

0.0762**

0.0405

0.0673*

0.0385

0.5045*** 0.0997* 0.1653

0.1173 0.0598 0.1182

-0.2321** -0.4508*** 0.2271*

0.1034 0.0856 0.1240

Number of Neighboring States

-0.0966**

0.0395

-0.0033

0.0420

2

0.0094** 3.8727 15,785 0.1505

0.0047 2.5462

0.0005 2.4392

0.0049 2.5312

(Number of Neighboring States) Constant Number of observation Pseudo R2

14,522 0.1432

*** significant at the 0.01 level , ** significant at the 0.05 level, *significant at the 0.1 level

Table 6 – Marginal Effects- Ordered Probit Model for Art Museum Attendance in 2008 Frequency Individual’s Characteristics Ln Real Income Number of Household Age Age*Age Female Disabled Education Education* Education Married Child Married*Child White Hispanic Employed Principal City* Employed Private

None

3 times

4-10 times

10>

-0.0682*** (0.0057) 0.0142*** (0.0031) -0.0019 (0.0012) 0.00004*** (0.0000) -0.0370*** (0.0064) 0.0813*** (0.0154) 0.157*** (0.0296) -0.0025*** (0.0004) 0.0435*** (0.0088) 0.0397** (0.0157) -0.0251 (0.0186) -0.0900*** (0.0106) 0.0186 (0.0119) 0.0340*** (0.0102) -0.0474*** (0.0183) 0.0156** (0.0076)

0.0542*** (0.0046) -0.0113*** (0.0025) 0.0015 (0.0010) -0.00003*** (0.0000) 0.0295*** (0.0051) -0.0669*** (0.0132) -0.125*** (0.0236) 0.0020*** (0.0003) -0.0344*** (0.0070) -0.0318** (0.0127) 0.0198 (0.0146) 0.0742*** (0.0091) -0.0149 (0.0096) -0.0269*** (0.0080) 0.0370*** (0.0140) -0.0124** (0.0060)

0.0123*** (0.0011) -0.0026*** (0.0006) 0.0003 (0.0002) -0.00001*** (0.0000) 0.0066*** (0.0012) -0.0127*** (0.0021) -0.0283*** (0.0055) 0.0005*** (0.0001) -0.0080*** (0.0017) -0.007** (0.0027) 0.0046 (0.0035) 0.0141*** (0.0015) -0.00326 (0.0020) -0.00624*** (0.0019) 0.0091*** (0.0037) -0.0028** (0.0014)

0.0017*** (0.0002) -0.0004*** (0.0001) 0.0001 (0.0000) -0.000001*** (0.0000) 0.0009*** (0.0002) -0.0016*** (0.0003) -0.0039*** (0.0009) 0.00006*** (0.0000) -0.0011*** (0.0003) -0.0009** (0.0004) 0.0007 (0.0005) 0.0017*** (0.0002) -0.00044 (0.0003) -0.0009*** (0.0003) 0.0013*** (0.0006) -0.0004** (0.0002)

-0.014 (0.0148)

0.0111 (0.0117)

0.0026 (0.0027)

0.0004 (0.0004)

Location Variables Principal City

68

Metropolitan West South Midwest Days of Snow Coast State Variables Percentage of Free Museums in the State Ln Real Income Level Ln Tax Revenue Per Capita Number of Neighboring States (Number of Neighboring States)2 N

-0.0273*** (0.0082) -0.0867*** (0.0131) 0.0163 (0.0131) -0.0167 (0.0129) -0.0008*** (0.0003) -0.0276** (0.0108)

0.0219*** (0.0066) 0.0671*** (0.0099) -0.013 (0.0105) 0.0132 (0.0102) 0.0007*** (0.0002) 0.0219** (0.0086)

0.0048*** (0.0014) 0.0170*** (0.0028) -0.0029 (0.0023) 0.00305 (0.0024) 0.0002*** (0.0000) 0.0050** (0.0020)

0.0006*** (0.0002) 0.0026*** (0.0005) -0.0004 (0.0003) 0.000429 (0.0003) 0.00002*** (0.0000) 0.0007** (0.0003)

-0.149*** (0.0315) -0.0364 (0.0315) -0.0286* (0.016) 0.0312** (0.0105)

0.118*** (0.0251) 0.0289 (0.0251) 0.0227* (0.0127) -0.0248** (0.0084)

0.0267*** (0.0057) 0.00654 (0.0057) 0.00514* (0.0029) -0.0056** (0.0019)

0.0037*** (0.0009) 0.0009 (0.0008) 0.0007* (0.0004) -0.0008** (0.0003)

-0.0030*

0.0024*

0.0005*

0.00007*

(0.0012)

(0.0010)

(0.0002)

(0.0000)

15,739

*** significant at the 0.01 level , ** significant at the 0.05 level, *significant at the 0.1 level

Table 7- Random Effect Model for Art Museums in 2002 and 2008. Model 1 Ln Attendance Museum Characteristics Age of the Museum/Gallery (Age of the Museum/Gallery)2 Museum University/College Private Museum/Gallery Free Admission Fee (Adult) (Admission Fee (Adult))2 Number of ulltime employee Museum Location Variables Ln Distance to the Biggest City Ln Distance to the Second Biggest City Number of Museums/Galleries with the Same Zip Code South

Model 2

Coefficient

Standard Error

0.0108*** -0.0000** 0.539*** -0.201** -0.183* -0.257***

(0.0028) (0.000) (0.0885) (0.093) (0.108) (0.0513)

Coefficient

Standard Error

0.0100*** -0.0000** 0.510*** -0.200** -0.192*

(0.0028) (0.000) (0.0877) (0.0916) (0.106) (0.0144) (0.0011) (0.0004)

0.0054***

(0.0004)

0.0794*** -0.0032** 0.0052***

-0.0630*** -0.0777***

(0.0198) (0.0228)

-0.0615** -0.0777***

(0.0195) (0.0225)

0.085* 0.692***

(0.0459) (0.153)

0.0769* 0.706***

(0.0456) (0.151)

69

Midwest West Coast Days of Snow

0.303** 0.508** -0.0722 0.0051**

(0.139) (0.174) (0.104) (0.0026)

0.318** 0.493** -0.085 0.0051**

(0.137) (0.171) (0.103) (0.0025)

Art Lovers Ln State Population Ln State Size

0.0057* 0.0871** 0.0363 7.798***

(0.0033) (0.0416) (0.0589) (0.621)

0.0075** 0.0881** 0.0369 7.511***

(0.0033) (0.0412) (0.0581) (0.612)

State Variables

Constant Number of observation R Square (between)

1,714 0.4330

1,714 0.4191

*** significant at the 0.01 level , ** significant at the 0.05 level, *significant at the 0.1 level

Table 8- State Characteristics Effect on Art Participation Standard Error Coefficient

Ln Average Attendance State Variables Art Lovers(%)

0.0702

(0.0035) (0.384)

Average Admission Fee (Adult)

0.0802**

(0.0392)

Ln National Park Visitors

0.0603**

(0.0226)

Ln Population

0.141**

(0.0453)

Percentage of Free Museums in the State

0.0081**

-0.0076

(0.0746)

Ln Tax Revenue Per Capita

0.302

(0.206)

Ln Real Income Per Capita

0.617*

(0.365)

Number of Neighbor States

0.0540**

(0.0255)

Most Visited States by Tourists Midwest West South Days of Snow

0.250** 0.620*** 0.352** 0.413*** 0.0055** -0.239** -2.055

(0.098)

Ln State Size

Location Variables

Year 2008 Constant

Number of Observation R Square

96 0.63

*** significant at the 0.01 level , ** significant at the 0.05 level, *significant at the 0.1 level

70

(0.121) (0.146) (0.117) (0.0026) (0.101) (4.568)

Figure 1-Total Appropriations to the National Endowments of Art and the State Agencies, 19702012

Source: “How the United States Funds the Art”, National Endowments of the Art. Note: Figures are in current dollars.

Figure 2- Price Elasticity of Demand for Art Museums- Public and Total Demand (by Selected State)

Price Elasticity of Demand for Art Museums Public and Total Demand 2 1.5 1 Elasticity (Total)

0.5

Elasticity (Public)

0 -0.5

NY

CA

WA MA

VA

TX

PA

NC

-1 -1.5 -2

71

NJ

FL

IL

NV

Figure 3-Price Elasticity of Demand for Art Museums – Public and Total Demand (by Region)

Price Elasticity of Demand for Art Museums Public and Total Demand (by Regions) 0.5 0 northeast

west

midwest

south

Elasticity (Total Atandence)

-0.5 Elasticity (Public Atandence)

-1 -1.5 -2

72

Appendix The Model 

The consumer utility function subject to budget constraint: (3.2) and (3.3)

max U  U ( A  A ,  , F ( F , ), E, Z ) s.t. M  PA A( AS , Ae , Ap , A , )  CF F ( FS , Fe , Fp , F , )  PE E ( E pr , Ec , Eov )  Pz Z 

The expenditure function subject to utility constraint: (3.2) and (3.3)

min M  PA A( AS , Ae , Ap , A , )  CF F ( FS , Fe , Fp ,F , )  PE E( E pr , Ec , Eov )  Pz Z s.t U ( A  A ,  , F ( F , ), E, Z )  U 

We include the condition of information to the utility function:

U  U ( A  A ,  , F ( F , ), E, Z | I E  1)

U  U (Z | I E  0) and

e  e( PA , CF , PE , Pz , A0 , F0 ,U 0 | I E  1)

e  e( Pz ,U | I E  0) 

Total Value= Use Value+ Existence Value: (3.6) and (3.7)



WTPChUsearg e = e( PA* , CF , PE , Pz , A0 , F0 ,U 0 | I E  1)  e( PA , CF , PE , Pz ,  A0 ,  F0 ,U 0 | I E  1) * 1 0 0 * 0 0 0 WTPChExistence arg e = e( PA , CF , PE , Pz ,  A ,  F , U | I E  1)  e( PA , CF , PE , Pz ,  A , F , U | I E 1)



Use WTPFree = e( PA , CF* , PE , Pz , A0 , F0 ,U 0 | I E  1)  e( PA , CF , PE , Pz ,  A0 , F0 ,U 0 | I E  1) Existance WTPFree = e( PA , CF* , PE , Pz ,  A , F1 ,U 0 | I E  1)  e( PA , CF* , PE , Pz , A0, F0 , U 0 | I E 1)

73

Results Table A1-Ordered Probit Model for Art Museum Attendance in 2008 and 2002 2008

Frequency Individual’s characteristics Ln real income Number of Household Age Age*Age Female Disabled Education Education* Education Married Child Married*Child White Hispanic Employee Principal City*Employee Private Sector Employee Individual’s Location Principal city Metropolitan West South Midwest Days of Snow Coast State Variables Percentage of Free Museums in the State Ln Tax Revenue Per Capita Ln Income Level

1 2 3

Coefficients

2002 Standard Errors

Coefficients

Standard Errors

0.2423*** -0.0504*** 0.0069 -0.0001*** 0.1320*** -0.3301*** -0.5585*** 0.0090*** -0.1528*** -0.1446** 0.0875 0.3675* -0.0676** -0.1192*** 0.1607*** -0.0555**

0.0205 0.0111 0.0044 0.0000 0.0229 0.0735 0.1043 0.0013 0.0306 0.0589 0.0639 0.0510 0.0444 0.0353 0.0595 0.0269

0.1812*** -0.0457*** 0.0110** -0.0002*** 0.1799*** -0.4517*** -0.3077*** 0.0058*** -0.0457 -0.1299** -0.0313 0.3487*** -0.1091** 0.0209 -0.0739 -0.1075***

0.0187 0.0109 0.0045 0.0000 0.0234 0.0812 0.1033 0.0013 0.0310 0.0540 0.0599 0.0492 0.0486 0.0355 0.0587 0.0275

0.0493 0.0993*** 0.2901*** -0.0585 0.0586 0.0030*** 0.0982**

0.0514 0.0303 0.0415 0.0475 0.0447 0.0010 0.0385

0.2244*** 0.1103*** 0.2547*** -0.0468 0.1072** 0.0027*** 0.0776**

0.0498 0.0301 0.0420 0.0483 0.0448 0.0010 0.0364

0.5285*** 0.1016* 0.1292

0.1120 0.0568 0.1119

-0.2261** -0.4083*** 0.2501**

0.0981 0.0807 0.1167

Number of Neighboring States

-0.1110***

0.0373

-0.0204

0.0395

(Number of Neighboring States)2

0.0106**

0.0044

0.0021

0.0046

-1.9712

2.4460

-0.3078

2.4202

-0.8393

2.4456

0.8078

2.4198

0.0361

2.4453

1.6367

2.4196

Number of observation 15,739 14,462 Pseudo R2 0.1227 0.1165 *** significant at the 0.01 level , ** significant at the 0.05 level, *significant at the 0.1 level

74

Table A2- Marginal Effects-Ordered Probit Model for Art Museum Attendance in 2002 Frequency Individual’s Characteristics Ln real income Number of household Age Age*Age Female Disabled Education Education* Education Married Child Married*Child White Hispanic Employed Principal City* Employed Private

None

3 times

4-10 times

10>

-0.0559*** (0.0058) 0.0141*** (0.0034) -0.0034** (0.0014) 0.00005*** (0.0000) -0.0551*** (0.0071) 0.117*** (0.0171) 0.0949*** (0.0320) -0.0018*** (0.0004) 0.0141 (0.0096) 0.0394** (0.0161) 0.0096 (0.0183) -0.0958*** (0.0118) 0.0325* (0.0140) -0.00644 (0.0109) 0.0223 (0.0174) 0.0332*** (0.0085)

0.0424*** (0.0044) -0.0107*** (0.0026) 0.0026** (0.0011) -0.00004*** (0.0000) 0.0419*** (0.0054) -0.0944*** (0.0145) -0.0720*** (0.0243) 0.0014*** (0.0003) -0.0107 (0.0073) -0.0301** (0.0124) -0.0073 (0.0139) 0.0758*** (0.0097) -0.0251* (0.0109) 0.00489 (0.0083) -0.0171 (0.0134) -0.0252*** (0.0065)

0.0114*** (0.0012) -0.00287*** (0.0007) 0.0007** (0.0003) -0.00001*** (0.0000) 0.0112*** (0.0015) -0.0200*** (0.0025) -0.0193*** (0.0066) 0.0004*** (0.0001) -0.00289 (0.0020) -0.00786** (0.0032) -0.0019 (0.0037) 0.0172*** (0.0019) -0.00635* (0.0026) 0.00131 (0.0022) -0.00443 (0.0034) -0.0068*** (0.0018)

0.00209*** (0.0003) -0.0005*** (0.0001) 0.0001** (0.0001) -0.000002*** (0.0000) 0.00205*** (0.0003) -0.00313*** (0.0004) -0.00355*** (0.0012) 0.00007*** (0.0000) -0.00053 (0.0004) -0.00142** (0.0006) -0.0004 (0.0007) 0.0028*** (0.0004) -0.00112* (0.0005) 0.00024 (0.0004) -0.00079 (0.0006) -0.0013*** (0.0003)

-0.0723*** (0.0167) -0.0334*** (0.0089) -0.0821*** (0.0141) 0.0143 (0.0147) -0.0337** (0.0143) -0.00083*** (0.0003) -0.0240** (0.0112)

0.0535*** (0.0120) 0.0255*** (0.0069) 0.0606*** (0.0101) -0.0109 (0.0112) 0.0253** (0.0107) 0.00063*** (0.0002) 0.0182** (0.0085)

0.0158*** (0.0039) 0.0066*** (0.0017) 0.0179*** (0.0033) -0.00289 (0.0029) 0.00703** (0.0031) 0.00017*** (0.0001) 0.0049** (0.0023)

0.00312*** (0.0009) 0.0012*** (0.0003) 0.00356*** (0.0008) -0.00053 (0.0005) 0.00133** (0.0006) 0.00003*** (0.0000) 0.0009** (0.0004)

0.0697**

-0.0529**

-0.0142**

-0.0026**

Location Variables Principal City Metropolitan West South Midwest Days of Snow Coast State Variables Percentage of Free Museums in the State

75

Ln Real Income Level Ln Tax Revenue Per Capita Number of Neighboring States (Number of Neighboring States)2 Number of observation

(0.0303) -0.0771** (0.0360) 0.126*** (0.0249) 0.0063 (0.0122)

(0.0230) 0.0585** (0.0273) -0.0956*** (0.0189) -0.0048 (0.0092)

(0.0062) 0.0157** (0.0074) -0.0256*** (0.0051) -0.0013 (0.0025)

(0.0012) 0.00289** (0.0014) -0.0047*** (0.0010) -0.0002 (0.0005)

-0.0007 (0.0014) 14,462

0.00049 (0.0011)

0.00013 (0.0003)

0.00002 (0.0001)

*** significant at the 0.01 level , ** significant at the 0.05 level, *significant at the 0.1 level

76

CHAPTER THREE ART AUCTIONS, PREMIUM RATES AND ART INVESTMENT

Abstract This article investigates whether and to what extent differing levels of a buyer’s premium affects the auction price of a painting. Using auction sales data on impressionist and contemporary paintings over the period 2000 and 2012, we show the relationship between the price of painting and its premium rates. We also show that how investment on art changes with market in London and New York. Examining final prices of paintings we see that paintings which have two possible premium rates tend to end up with high premium rates and low hammer prices (highest price). Moreover, an investment on impressionist and contemporary art between 2000 and 2012 in New York is relatively less risky than the market. However, there is no evidence that this is the case for an investment on impressionist and contemporary art in London

JEL Classification: Z11, D44, D11, D12

77

1. Introduction Buying and selling art is one of the most important financial decisions for art collectors all around the world especially for countries where art is a large and growing market. Art transactions vary in many ways. An art can be sold directly by its owner and other private individuals or by galleries and auction houses. Art transactions via galleries and auction houses help match sellers’ offerings and buyers’ demands. By this way, galleries and auction houses bear the cost of marketing the art piece (catalogs, advertisements). Galleries and auction houses usually work in different ways. For the service they provide, galleries and auction houses charge either fixed percentage commissions (galleries) or varying commissions to different price levels (auction houses) to sellers. Auction houses also charge to commission rates to buyers. Who gains or losses from the strategy of charging different buyer premium rates for bids at different price levels? This paper investigates the paintings of which prices are close to the limit for next premium rate or previous premium rate. In other words, we analyze the strategy of bidders for a painting when the hammer price reaches to the limit that is set by the auction house. Since the characteristics of a painting impact the painting’s hammer price and also its premium rate, we analyze the premium rates for paintings by using hedonic regression model. The role of art market in global economy is small but it is growing. Buyers consider if they can gain from art as an investment. Investment on art has been a topic in literature. Most of the existing literature on art investment is concerned with European, American, Canadian or Asian paintings. In this paper, we also analyze the valuation of impressionist and contemporary painting sold at Christie’s and Sotheby’s in London and New York over the period 2000 and 2012. By using the stock market for two countries and risk free assets, we estimate the capital

78

asset pricing model (CAPM). The results showed movements and risk of art market in United Kingdom and United States. The organization of this paper is as follows. First, we summarize art auctions with the characteristics of leader art auctions in the art market. After brief information about the leaders, we show the strategies of auction houses for their buyer’s premium rates and buyers’ behaviors towards premium rates for some specific price ranges by depending on hammer price limits. Last, we analyze the art market to show the risk of investment on paintings at those art auction houses.

2. Art Auctions In the art market, auctions for paintings are English auction type. That is, paintings are sold to bidders who offer the highest prices. Before the auction, the artwork is examined by experts. The estimated prices are announced in the pre-auction catalog. When price for the painting reaches the highest amount, auction stops and the highest price is called the “hammer price”. Not all the art work at auction might be sold. The final bid for some art works do not reach the seller’s pre-set reservation price. In these cases, the painting is called “bought in” and withdrew from the auction. Ashenfelter and Graddy (2003) state that auction system provides a public report on the price of art. Auctions also provide data to understand the mechanism of auctions and the buyer’s behavior. The art market returns can be calculated with the paintings at auction. Chanel, Varet and Ginsburgh (1996) suggested that price indices of paintings should be based on regressions using sold paintings, not only resold paintings. In this study, we use auction data to analyze the behavior of buyers and include sold paintings to calculate the returns.

79

The rate of return of art also was discussed in the literature. Anderson (1974) examined auction prices by using the repeat-sales method; he calculated the annual real rate of return and found that modern works such as impressionists and Twentieth century paintings’ auction prices increased at a higher rate than other schools. Buelens and Ginsburgh (1993) found that investing in paintings provided higher returns than bonds. The capital asset pricing model (CAPM) is another approach used for the art investment. Stein (1977) is the first to apply the CAPM to art market data (auction) for the 1946-1968 periods. He calculated the financial returns to the paintings auctioned in the US and the UK between 1946 and 1968. Stein’s results indicated that the returns to paintings are lower than stocks and bonds. We use CAPM in this paper and consider the time-series representation of CAPM developed by Jensen (1968). The most of the analyses about art prices and investment on art are based on art auctions. Two auction houses lead the art market, Sotheby’s and Christie’s. They have the power of the market among others such as Heritage Auctions, Phillips de Pury & Company and Bonhams. We focus on these two leaders, Sotheby’s and Christie’s, to address our question about the commission rates and art investment. 2.1 Christie’s and Sotheby’s Auction Houses Christie’s Auction House was founded in 1766 and holds global auction and private sales. The sale categories include jewelry, photography, wine and all types of fine art and decorative arts. Christie’s has 53 offices in 32 countries including in the United Kingdom, U.S.A., France, Switzerland, Italy, Netherlands and China and offers more than 450 sales annually.1 Christie’s offers five bidding options: in person, Christie’s LIVE, written bid, absentee bid online and telephone bid. Prices of art works range depends on the artists and the sale

1

th

http://www.christies.com/about-us/company/overview/. (access on February 20 , 2013)

80

category. Prices start from $200 to $100 million.2 In 2010, Christie’s announced that they had the record sale ($5 billion) in 2010 and it was the best year in its history with a 66% increase in dollar over 2009.

3

The total sale reported in 2011 was $5.7 billion with a 14% increase

compared with 20104 and $6.27 billion in 2012 with a 10% increase on 2011.5 Sotheby’s was founded in London in 1744. The presence of Sotheby’s in the world is larger than that of Christie’s. Sotheby’s has 90 locations in 40 countries with 250 auctions yearly in over 70 categories.6 The sale categories include antiques, old master paintings, impressionist and contemporary art, jewelry, furniture, watches and wine. Sotheby’s and Christie’s auctions are open to public. Similar to Christie’s auctions, bidders may be present in the saleroom, on the phone or online (BIDnow). Auctioneers also place bids on behalf of absentee bidders. In 2011, Sotheby’s announced their total sales as $5.8 billion, up from $4.8 billion in 2010. In 2012, the total sales decreased to $4.4 billion. The highest total annual sales for Sotheby’s in its history were $6.2 billion in 2007. 7 Figure 1 and Figure 2 show the market share for Christie’s, Sotheby’s and others from 2008 to 2011. When analyzing the development of the Christie’s and Sotheby’s art auctions by lots and turnover in the years from 2008 until 2011, it is obvious that there has been a continuous decrease in the size of the market share for Christie’s and Sotheby’s and increase in the other auction houses’ market share. Christie’s and Sotheby’s together represented 47% of global art sales in 2011. In 2008, they had over 73% of global art auction revenue. The competition among 2

th

http://www.christies.com/about/press-center/releases/pressrelease.aspx?pressreleaseid=5966 (last access on February 20 , 2013) 3 http://www.telegraph.co.uk/finance/newsbysector/retailandconsumer/8287252/Christies-auction-house-has-best-year-inth 245-year-history.html (last access on February 20 , 2013) 4 th http://www.christies.com/about/press-center/releases/pressrelease.aspx?pressreleaseid=5356 (last access on February 20 , 2013) 5 th http://www.christies.com/presscenter/pdf/2013/RELEASE_2012_CHRISTIES_RESULTS.pdf (last access on February 20 , 2013) 6 nd http://www.sothebys.com/en/inside/about-us.html (last access on February 22 , 2013) 7 nd http://www.artnet.com/magazineus/news/artmarketwatch/sothebys-2010-results-3-1-12.asp(last access on February 22 , 2013)

81

auction houses in the art market has been growing. Although Christie’s and Sotheby’s have been losing the power against the other auction houses, they are still the two leaders of the market. The market share of turnovers illustrates how big the transactions are at auction houses. Because of the high market shares for Christie’s and Sotheby’s, it is evident that the total amount of the transaction (in dollars) at these auction houses is significant compared with other auction houses. The most expensive paintings sold at auctions are usually via Christie’s or Sotheby’s auction houses. 8 Figure 2 demonstrates the market trends based on share of lots sold. The difference in lots sold between other auction houses and Christie’s and Sotheby’s is greater than that based on turnovers. This shows how the value going through Christie’s and Sotheby’s is large, but the volume is low. The most valuable and expensive art works are sold by the leaders of the market even though they do not have very big impact on the number of lots sold. Other auction houses include China Guardian (2% lots sold, 8% turnovers) and Poly International (1% and 8% turnovers) in 2011. China Guardian was found 1994 and today it is the third largest auction house in the world.9 This is due to the growth of Chinese art and its impact on the global art market. 2.2 Premium Rates Auction houses income come from commissions, fees charged to clients for catalogue production and insurance, catalogue subscription revenues, advertising revenues and

8

For example,” Number 5, 1948” from Jackson Pollock was sold in Sotheby’s private sale for $140 million in 2006, making it the most expensive painting ever sold (up to now). Christie’s sold Gustav Klimt's "Adele Bloch-Bauer I." for $135 million at the same year (2006). It became the second most expensive painting ever sold at auctions. 9 Art Market Trends 2011, artprice.com

82

commissions earned on private sales brokered by third-parties.10 However, their income is mostly from the commission rates charged to sellers and buyers. 11 An auction house charges a commission for a selling art work. The commission includes those paid by the buyers and those paid by the sellers. The commission charged to a buyer is called “buyer’s premium” and to a seller is “seller’s commission”. The commissions from both buyers and sellers are calculated as a percentage of the hammer price of the art work sold at auction. The hammer price plus the buyer’s premium is called the “buyer’s price” or “final price,” while the hammer price less the seller’s commission is called the “seller’s price.” Auction houses also specify buy-in penalties, in addition to commission rates, in their standard auction contracts. The seller has to pay fee to the auction house if the property fails to be sold. By doing this, auction houses also generate income from for the unsold paintings (not only for sold paintings). Greenleaf and Sinha (1996) investigated why buy-in penalties exist and examined how auction houses can use buy-in penalties most effectively. They found that implementing penalties on unsold paintings motivates the seller to set a lower reserve and increased the total expected auction revenues. On the other hand, art auction may guarantee for some pieces. Because of the high competition between Sotheby’s and Christie’s, the auction houses have some strategies to attract sellers. The guarantee arrangement includes the guarantee amount and the extra commission the seller pays if the auction price exceeds the guarantee. Guarantees increase the expected value and benefits to sellers. However, they decrease the expected revenue for auction houses (Greenleaf, Rao, Sinha 1994). After March 1995, this

10

Sotheby’s Annual Report, 2011 Sotheby’s announced that action commission revenue accounted 84%, 86%, and 82% of the total revenue in 2011, 2010 and 2009, respectively. 11

83

competition ended. Christie’s and Sotheby’s announced that they would charge fixed and negotiable commission to sellers for the sold pieces (Ashenfelter and Graddy 2004). The calculation of commissions and premiums depends on hammer price. The rate decided by auction houses is called “commission rate” for sellers and “premium rate” for buyers. Buyers’ premiums of 10% were first implemented in 1975 by Christie’s and later by Sotheby’s (Ashenfelter and Graddy 2004). The commission rate method has changed since its initial implementation.

Nowadays, there is a negative relationship between commission rate and

hammer price. Thus, the commission rate applied on artwork increases while the hammer price decreases. Table 1 shows the buyer‘s premium for the time period of 2008- 2013. The premium rate charged to buyers for an artwork up to $50,000 (hammer price) is greater than for an artwork over $1,000,000 (hammer price). The premium rate decreases as the hammer prices increases. Auction houses charge higher percentage for smaller amounts. During 2000-2012 time periods, premium rates changed several times for auctions in New York and London. The changes were either in premium rates or in minimum/maximum limits. Table 2 and Table 3 show the dates of changes in premium rates both for Sotheby’s and Christie’s in New York.12 It is clear from the dates that the auction houses change the premium rates after another or at the same time. Although this movement between auction houses brought some evidence for establishment of the price-fixing agreement in 2001, Sotheby’s denied the presence of collusion respect to fixing buyers’ premiums (Ashenfelter and Graddy, 2004).

12

See Appendix for the buyer’s premium in London.

84

2.3

Collusion-Price Fixing for Auction Houses

The seller’s commission rate has also changed over time. In March 1995, Christie’s decided to have a fixed non-negotiable commission rate for sellers effective from September 1995. In April 1995, Sotheby’s announced that they would have the same policy. This created some problems. In June 1996, the UK Office of Fair Trading claimed that there are possible anticompetitive practices at Sotheby’s and Christie’s. In addition to issues with the seller’s commission, Christie’s and Sotheby’s are also considered for price fixing to buyers starting from 1993. In short, anyone who had bought items in the United States from Christie’s or Sotheby’s between 1 January 1993 and 7 February 2000 and anyone who had sold items between 1 September 1995 and 7 February 2000 were involved in this case. In 2001, while Sotheby’s as a company paid the $45 million criminal fine and the $256 million fine for damages to the buyers and the sellers, Christie’s wasn’t charged with a criminal fine because of their cooperation with the Justice Department but had to pay the same amount as Sotheby’s to the buyers and the sellers. Ashenfelter and Graddy (2004) discussed this problem from the buyers’ side. They pointed out their previous study (Ashenfelter and Graddy, 2003). They studied the theory of private value auctions which indicates that the higher the premium rates, the more the buyers reduce their bids. In other words, the total increase in the buyers’ premiums should fall on the seller. They claim that the price-fixing would affect the buyers. However, the theory would work under the assumption of fixed number of sellers and buyers and no seller reservation price. Ginsburgh, Legros and Sahuguet (2005) took into account the reservation price and concluded that the welfare of all bidders is the same, regardless of the commission. Ashenfelter and Graddy

85

(2004) summarized that the impact of increasing in commission rate should be very small and fall on the sellers. Sellers were injured the most from the collision. They paid higher commission because of the price-fixing. Also, some sellers withdrew their items from the market. Our concern is not to discuss about the collusion between Christie’s and Sotheby’s. However, we analyze the impact of setting same price limits and premium rates on the rate of premium returns over time periods when premium rates changed. We also focus on buyers’ strategies for the different premium rates to different price levels at one auction.

3. Theoretical Implication Ashenfelter and Graddy (2004) pointed out that when buyers’ premiums are implemented to auctions, each buyer will reduce his reservation price by an equivalent amount. Thus, the seller’s revenue will decrease by the amount of the buyers’ premium. If there is premium rate (c) for buyers, then buyers will reduce his reservation price ( v1 ) to compromise the commission. Then, the reservation price of the buyer will be v1  p1'  p1' * c where p1'  p1 . The hammer price is

v2 . 1 c

In our study, we discuss the situation when the highest bid comes close to the limit (m) for the premium rates. Then, bidders should consider the next premium rate, which is lower (cl) than the previous premium rate (ch), ch > cl. In that case, the bidder will bid higher than the limit(m) and hammer price would be

v2 v v instead of 2 .His reservation price is p1'  1 . 1  cl 1  ch 1  cl

86

This would be possible only if m-e ≤

v2 v ≤ m and 2 ≥ m+e where e is a small number. The 1  ch 1  cl

final price would be same, however, the hammer price would be higher and the premium to the auction house would be lower. Who wins or loses? Assume that the seller pay a fix commission rate (f), the commission they would pay for high hammer price would be high and for low hammer price it would be low. Obviously, auction house loses from the buyer premium because e is a small number. The auction house loses because

cl  f ch  f when 0