John Page
A. B. Freeman School of Business

Thomas Bayer
Department of Art History

Tulane University

Plan of presentation

• Descriptive tatistics and characteristics of the art market
• Development of an art market index
• Hedonics: estimation of prices for artworks from their characteristics
• Toward a model of the market, giving characteristics of market participants
• Future steps in this research

Primary source:

Graves, Algernon. Art sales from early in the eighteenth century to early in the twentieth century. London, 1918-21; rpt. New York: B. Franklin, 1970. 3 vols.

Graves Data for Each Transaction

• Artist name, with dates of birth and death
• Picture title
• Picture medium, date, height, and width
• Auction house, with its lot number
• Buyer name
• Seller name
• Price in pounds, shillings, and pence

Art Market Transactions, 1740-1909

Description	Number
Total number of transactions	39,568
Number with artist identified	39,457
Number with picture title given	39,552
Number with buyer name	30,147
Number with seller name	34,008
Number with auction house	39,567

Counts of the Art Market Transactions

Description	Number
Number of artists	1,837
Number of separate artworks	33,266
Number of buyers	3,551
Number of sellers	3,890
Number of auction houses	90
Number of picture mediums	4

Overall Statistics for the Transactions

measurement	N	Mean	Median	Min	Max	Sum
Price (£)	34,797	327	158	0.4	24,250	11,380,004
Length (in.)	17,057	30	25	1	303
Width (in.)	17,058	29	25	1	374
Size (in.²)	17,055	1,115	650	1	54,540
Picture date	5,481	1841	1862	1463	1907
Sale date	39,124	Mar1868	May1881	Mar1741	Dec1909

Coding added to Graves data

• Art code for artists
  • Contemporary vs. Old Master
  • Continental vs. English
• Buyer status
• Seller status
• Artist, buyer, and seller sex
• Subject of artwork

Statistics for Art Code

Art Code	N	Nₚ	Median price (£)	Nₛ	Median size	Total value (£)
Contemporary Continental	3,340	3,281	210	2,217	391	1,216,538
Contemporary English	15,227	14,539	179	7,357	600	4,866,248
Old Master Continental	11,703	8,898	168	4,035	644	3,064,499
Old Master English	9,187	7,970	74	3,369	750	2,210,008

Old Master Painters

Continental

• Rubens — 536
• Rembrandt — 392
• David Teniers — 344
• B.E. Murillo — 282
• Jacob Ruysdael — 261
• Albert Cuyp — 234
• P. Wouvermans — 231
• VanDyck — 220

English

• Joshua Reynolds — 2,423
• Gainsborough — 1,490
• George Romney — 515
• George Morland — 423
• Thomas Lawrence — 294
• John Constable — 273

Contemporary Painters

Continental

• Birket Foster — 252
• Peter De Wint — 183
• Josef Israels — 152
• Pierre Frere — 143
• Rosa Bonheur — 135
• Verboekhoven — 132
• Camille Corot — 106
• Fantin-Latour — 91

English

• J.M.W. Turner — 1,413
• David Cox — 657
• Edwin Landseer — 648
• Thomas Cooper — 552
• John Linnell — 485
• William Hunt — 388
• Anthony Fielding — 339
• William Muller — 309

Statistics for Types of Buyers

Buyer	N	Nₚ	Median price (£)	Nₛ	Median size	Total value (£)
Artist	47	47	105	6	759	9,701
Bought-in	2,182	2,150	149	288	1,033	602,980
Dealer	16,727	16,686	205	9,766	587	7,069,830
Institution	85	85	525	21	520	88,296
Middle class	19,732	15,039	116	6,775	696	3,346,864
Upper class	795	790	151	199	750	262,334

Statistics for Types of Sellers

Seller	N	Nₚ	Median price (£)	Nₛ	Median size	Total value (£)
Artist	2,903	2,112	39	637	690	245,546
Dealer	1,478	1,311	142	611	1,008	341,866
Institution	176	71	147	28	399	14,132
Middle class	29,064	26,133	168	13,351	612	8,622,283
Upper class	5,947	5,170	168	2,428	736	2,156,177

Sex Effects

Category	N	Nₚ	Median price (£)	Nₛ	Median size	Total value (£)
Artist, Female	350	301	210	178	645	147,980
Artist, Male	39,218	34,496	158	16,877	650	11,232,024
Buyer, Female	51	51	100	18	831	9,803
Buyer, Male	39,517	34,746	158	17,037	648	11,370,200
Seller, Female	1,746	1,653	137	655	612	528,059
Seller, Male	37,822	33,144	161	16,400	655	10,851,945

Statistics for Art Subjects

Subject	N	Nₚ	Median price (£)	Nₛ	Median size	Total value (£)
Animal	2,290	2,124	158	1,023	713	635,131
Genre	8,065	7,275	179	3,638	546	2,395,497
History	1,547	1,326	152	550	1,008	414,657
Interior	356	311	158	157	374	102,611
Landscape	12,519	11,614	179	6,015	486	3,884,493
Marine	1,210	1,125	194	548	570	381,159
Mythology	1,470	1,024	142	423	1,160	339,545
Portrait	7,481	6,533	105	3,100	750	2,153,356
Religious	4,045	2,942	142	1,274	832	959,368
Still Life	576	523	158	327	252	114,188

Reactions to summary statistics

Characteristics of art works change

• Old master English paintings have more portraits
• Contemporary paintings have more landscapes; coincidentally(?), landscapes sell for the highest price per square inch
• Contemporary English have more animal, history, and marine
• Contemporary continental paintings are smaller

Are these changes the result of dealers?

Characteristics of the Art Market

“Bad”

• Transactions occur at unevenly spaced dates
• Transactions are usually not of the same item

“Good”

• Have transactions over 170 years
• Artists often resold, even if individual works are not
• Frequently know both buyer and seller

Calculation of art market index

• Cannot use mean or median prices because of potential changes in buying and selling patterns
  • Change in average quality or size
  • Buy-ins around real price decreases
  • Change in type of art purchased (taste)
• We will use artwork actually sold more than once during the time period

Identifying resales of artworks

• Begin with artist match and title match, but
  • Title given in many forms, different languages
  • Same title given to preliminary drawings, smaller copies, and similar paintings
  • Picture date often missing, size given inconsistently
• Tried to match on picture date, picture medium, and size
• Presumption was artworks sold less than 5 times were the same, more than 5 times were different
• Sample size goes to 2,090

Adaptation of Housing Index Regression

Source: J.A.S.A., Bailey, Muth and Nourse

• yᵢ = log₁₀(price of the iᵗʰ artwork at second sale)
  • log₁₀(price on its first sale date)
• Independent variables consist only of dummy variables, one for each time period in the sample except for the first.
• For each artwork, the dummy variables are zero except for the dummy corresponding to the second sale, where it is +1, and for the dummy corresponding to the first sale, where it is -1.
• The estimated coefficients are then taken as the log price index.

Estimation of price: hedonics

• Begin with estimate of an individual artist
• Estimate price per square inch
• Characteristics include
  • Artist living or dead
  • Picture medium
  • Subject of painting
• For regression of all artists distinguish
  • Contemporary vs. Old Master
  • Continental vs. English

Analysis of Variance of COX

Source	DF	Sum of Squares	Mean Square	F Value	Prob > F
Model	9	101.60145	11.28905	16.900	0.0001
Error	365	243.81541	0.66799
C Total	374	345.41686

Statistic	Value	Statistic	Value
Root MSE	0.81730	R-square	0.2941
Dep Mean	-0.22285	Adj R-sq	0.2767
C.V.	-366.74668

Parameter Estimates

Variable	DF	Parameter Estimate	Standard Error	T for H0: Parameter=0	Prob > |T|
INTERCEP	1	-13.332847	6.64107410	-2.008	0.0454
LIVING	1	-2.271032	0.26880794	-8.449	0.0001
MEDIUM	1	-0.203796	0.09077196	-2.245	0.0254
Y	1	0.007082	0.00350630	2.020	0.0441
ANIMAL	1	-0.266431	0.21808228	-1.22	0.2226
RELIGION	1	0.319806	0.39422345	0.811	0.4178
MARINE	1	0.007702	0.3427597	0.022	0.9821
PORTRAIT	1	-1.042711	0.43811836	-2.380	0.0178
LANDSCAPE	1	-0.114781	0.15540833	-0.739	0.4606
STILLIFE	1	-0.054919	0.83107343	-0.066	0.9473

GARCH Estimates for COX

Statistic	Value	Statistic	Value
SSE	218.226	OBS	370
MSE	0.5898	UVAR	0.590282
Log L	-426.952	Total Rsq	0.3682
SBC	948.5204	AIC	885.9043
Normality Test	8.5978	Prob > Chi-Sq	0.0136

Variable	DF	B Value	Std Error	t Ratio	Approx Prob
Intercept	1	0.063838	0.1678	0.380	0.7036
LIVING	1	-2.475227	0.3880	-6.379	0.0001
MEDIUM	1	-0.197508	0.1023	-1.930	0.0536
PORTRAIT	1	-0.583298	0.3867	-1.508	0.1315
A(1)	1	-0.251741	0.0557	-4.522	0.0001
A(2)	1	-0.093673	0.0581	-1.611	0.1071
ARCH0	1	0.559511	0.0558	10.029	0.0001
ARCH1	1	0.052129	0.0699	0.746	0.4558
GARCH1	1	6.059098E-23	3.29E-15	0.000	1.0000

Ordinary Least Squares Estimates for All Artists

Statistic	Value	Statistic	Value
SSE	28670.62	DFE	16402
MSE	1.747996	Root MSE	1.322118
SBC	55888.41	AIC	55772.82
Reg Rsq	0.1463	Total Rsq	0.1463
Durbin-Watson	1.2906

Variable	DF	B Value	Std Error	t Ratio	Approx Prob
Intercept	1	-26.543054	0.9189	-28.886	0.0001
LIVING	1	-0.074038	0.0309	-2.397	0.0166
MEDIUM	1	0.816703	0.0347	23.507	0.0001
Y	1	0.013660	0.000488	28.000	0.0001
ANIMAL	1	-0.271842	0.0475	-5.728	0.0001
RELIGION	1	-0.478679	0.0470	-10.194	0.0001
MYTH	1	-0.537583	0.0711	-7.566	0.0001
MARINE	1	-0.181079	0.0611	-2.961	0.0031
INTERIOR	1	0.052770	0.1098	0.480	0.6309
PORTRAIT	1	-0.301074	0.0379	-7.948	0.0001
LANDSCAPE	1	0.008559	0.0290	0.295	0.7680
STILLIFE	1	0.150821	0.0778	1.939	0.0525
HISTORY	1	-0.576309	0.0623	-9.255	0.0001
CONTEMP	1	0.232129	0.0292	7.957	0.0001
ENGLISH	1	-0.553530	0.0234	-23.628	0.0001

GARCH Estimates for All Artists

Statistic	Value	Statistic	Value
SSE	18983.72	OBS	12955
MSE	1.465359	UVAR	1.476947
Log L	-20716	Total Rsq	0.4347
SBC	41782.39	AIC	41506.03
Normality Test	152.6979	Prob > Chi-Sq	0.0001

Variable	DF	B Value	Std Error	t Ratio	Approx Prob
Intercept	1	-0.583127	0.0425	-13.724	0.0001
LIVING	1	-0.063857	0.0388	-1.646	0.0997
MEDIUM	1	0.642079	0.0463	13.859	0.0001
ANIMAL	1	-0.207524	0.0505	-4.113	0.0001
RELIGION	1	-0.408971	0.0449	-9.119	0.0001
MYTH	1	-0.353767	0.0672	-5.268	0.0001
MARINE	1	-0.126679	0.0641	-1.976	0.0481
INTERIOR	1	-0.008182	0.1059	-0.077	0.9384
PORTRAIT	1	-0.204754	0.0353	-5.794	0.0001
LANDSCAPE	1	-0.020422	0.0302	-0.676	0.4990
STILLIFE	1	0.033685	0.0914	0.369	0.7123
HISTORY	1	-0.422519	0.0553	-7.637	0.0001
CONTEMP	1	0.048916	0.0334	1.463	0.1434
ENGLISH	1	-0.472977	0.0285	-16.615	0.0001
A(1)	1	-0.290487	0.00947	-30.687	0.0001
A(2)	1	-0.095721	0.00998	-9.595	0.0001
—
A(20)	1	-0.018871	0.00864	-2.185	0.0289
ARCH0	1	0.309326	0.0347	8.923	0.0001
ARCH1	1	0.102353	0.00916	11.176	0.0001
GARCH1	1	0.688211	0.0285	24.172	0.0001

Results of hedonic regressions

• Different artists have statistically different prices
• “Death is good.”
• Landscape, still life, and interior paintings have highest price per square inch (measure of effort?)
• Religious, mythological, and history paintings sell for less
• English artists sell for less
• Strong autocorrelation of prices

Market participants

• Buyers. Early, dealers are not buyers, but by 1870 buyers are primarily dealers.
• Sellers. These are primarily infrequent sellers, usually middle class or upper class, often at death. Dealers rarely sell.
• Auction houses. Christie’s does not start until 1770, but quickly becomes dominant.

TABLE OF LARGEST BUYERS BY ARTCODE

Frequency / Row Pct

Buyer	CC	CE	OMC	OME	Total
Agnew	415 8.28	3,323 66.29	456 9.10	819 16.34	5,013
Colnaghi	34 3.18	144 13.46	564 52.71	328 30.65	1,070
McLean	114 15.30	481 64.56	25 3.36	125 16.78	745
Tooth	240 29.27	471 57.44	27 3.29	82 10.00	820
Vokins	130 13.21	705 71.65	39 3.96	110 11.18	984
Wallis	250 35.92	289 41.52	44 6.32	113 16.24	696
Total	1,183	5,413	1,155	1,577	9,328

Support of artists by dealers

Most

• Maris — 84%
• Frere — 80%
• Graham — 76%
• Guardi — 76%
• Alma-Tadema — 73%
• Davis — 71%
• Collins — 70%
• Josef Israels — 70%

Least

• Carracci — 29%
• Raphael — 26%
• Rossetti — 26%
• Guido Reni — 25%
• Velasquez — 24%
• Titian — 19%
• Poussin — 15%
• Guercino — 3%

TABLE OF LARGE SELLERS BY ARTCODE

Frequency / Row Pct

Seller	CC	CE	OMC	OME	Total
Mr. Bryan	0 0.00	0 0.00	203 98.54	3 1.46	206
A.C. Calonne	0 0.00	0 0.00	217 95.59	10 4.41	227
Hamilton Palace	0 0.00	2 0.98	187 91.22	16 7.80	205
H.J.A. Monro	0 0.00	125 54.35	74 32.17	31 13.48	230
Messrs. Murietta	76 35.85	113 53.30	17 8.02	6 2.83	212
J.H.S. Pigott	0 0.00	5 2.39	111 53.11	93 44.50	209
Sir J. Reynolds	0 0.00	0 0.00	286 96.62	10 3.38	296
Benjamin West	0 0.00	0 0.00	70 27.34	186 72.66	256
Total	76	245	1,165	355	1,841

Number of times at Auction

Buyers

• Agnew — 748
• Colnaghi — 404
• Vokins — 369
• Tooth — 364
• McLean — 350
• Wallis — 303
• Permain — 203
• Gooden — 195

Sellers

• Henry Wallis — 25
• Thomas McLean — 17
• Woodburn — 12
• Alton — 11
• Mr. Bryan — 11
• Colnaghi — 11
• L.V. Flatou — 10
• Hamilton Palace — 10

Dealer market characteristics

• Dominate as buyers
• Pay more on average
• Infrequently use auctions to sell
• Specialize by art code and artist (To avoid competition?)
• Avoid
  • Lower-priced artwork
  • Larger paintings
  • Religious and mythological paintings

Characteristics of non-dealers

• Sell very infrequently
• Artists purchase in lower-price range
• Highest buy-in rate is in lowest price range
• Institutions concentrate on higher-priced paintings
• Highest buy-in rate is in largest paintings

Economic model of art market

• Dealer appears to function as market maker
• Dealer can pay more because of his inventory
• Sellers are liquidity traders
• Evidence of efficiency
  • Cannot buy and sell profitably in the auction market
  • Artwork bought-in sells for less than the buy-in price
  • Highest buy-in rate is for dealers as sellers
  • Prices rise after death
• Evidence of inefficiency: strong autocorrelation

What needs to be done

• Reduction of dimensionality in hedonic regression by grouping statistically similar artists together
• Interpret Arch/Garch results
• Decide whether this database and concepts should be one paper or two
• Develop economic model of dealers and this market

Possible future research questions

• Would the use of hedonic regressions help produce a better art index?
• How would these results change or remain consistent with current data?
• What is the implicit bid/ask spread of the dealer? Is it currently decreasing?
• Is the traditional art market now breaking down?
• Were there (Are there) opportunities to buy profitably?
• Does a change in styles help artist prices?