The labor market in the art sector of Baroque Rome.
Etro, Federico ; Marchesi, Silvia ; Pagani, Laura 等
"Caravaggio said that it takes as much manufacture to do a
good painting of flowers as of human figures." Vincenzo
Giustiniani, Letter on painting (1620-1630)
I. INTRODUCTION
A wide literature in cultural economics, surveyed by Ashenfelter
and Graddy (2003), has analyzed the pricing of art through hedonic
regressions based on auction data. This analysis is focused on secondary
markets and sheds light on the determinants of art value, the life cycle
of painters (Galenson and Weinberg 2000), or the sources of artistic
success (Hellmanzik 2010). (1) Similar techniques can be applied to
historical primary markets to shed light on the determinants of the
earnings of the artists and of the characteristics of their commissions.
Only recently, however, art historical studies have put together wide
datasets on prices of the artistic commissions during the Renaissance
and the Baroque age (see De Marchi and Van Miegroet 2006; O'Malley
2005; Spear and Sohm 2010) and economic investigations are now possible.
(2) In this article, we analyze the seventeenth-century market for
painters based in Rome through a unique dataset on paintings of the
Baroque age, which allows us to focus on the (occupational) choice of
the artistic genre by painters and on equilibrium pricing across art
sectors.
The most impressive and rapid phenomenon of the seventeenth-century
art industry has been the innovative form of artistic differentiation
that led to the mass production of new genres of paintings. Besides
traditional figurative paintings, mainly on religious, mythological, and
historical subjects, and besides portraits, the new genres of the
Baroque art market included what were considered "minor
genres" such as still lifes (reproducing animals, fruits, flowers,
and lifeless objects), landscapes (reproducing the countryside or the
urban environment), so-called genre paintings (reproducing daily life
scenes), and battles (reproducing fights without necessarily a specific
historical content). Each genre represented a specific sector of
production (with limited substitutability between genres on the demand
side), and painters either specialized in one or few genres or they
could switch between them according to market opportunities.
The prestige of the genres was rigorously ranked in the artistic
culture of the time, and the most dignified and worthy subjects were
those depicting creative compositions of idealized human figures, as in
religious and mythological paintings. Behind these, a relative
preference was reserved for landscape paintings, which had to represent
idealized views of the world. Compositions of the daily aspects of
reality, the so-called genre paintings, were at a lower level in the
ranking of the art commissioners, while the least worthy genres were
those imitating reality without idealization like portraits and, at the
lowest level, still lifes. (3) Such a ranking of preferences for the
genres was well understood between art collectors, art critics, and
artists, and later codified by the art academies. A preliminary view of
the data on primary sales of paintings suggests that this hierarchy was
also associated with a clear ranking of payments among the different
genres. Price differentials between them were sizable: the average
prices in Roman silver scudi were, respectively, 17 scudi for still
lifes, 25 for genre paintings, 39 for portraits, 66 for landscapes, and
240 for all the figurative paintings (with further differences between
subcategories). Paintings of different genres obviously differed in many
respects (e.g., size, technique, support, and destination) and price
differentials did not necessarily reflect differences in the effective
compensation of the painters. However, as long as the market was
competitive and painters could freely choose whether to specialize in
one genre 3 or switch between genres to exploit profitable
opportunities, we can hardly imagine that systematic compensation
differentials could persists between artistic sectors. In other words,
in equilibrium, painters and paintings should have been allocated
between different commissions to the point of equalizing the marginal
return of each genre.
The main objective is to test this hypothesis of price equalization
between artistic sectors. To do so, we adopt a labor market framework in
which painters can be thought as workers, patrons can be thought as
employers, and genres can be interpreted as distinct artistic sectors.
Rather than wages (e.g., annual or hourly wages, as in the standard
labor market literature) the data report the price of each single item
(painting) produced by each artist. In labor economic terms, we would
define the compensation for each painting (i.e., the painting price) as
a compensation at a piece rate rather than at a time rate. If
artists' mobility between artistic fields (genres) was free, we
expect that, after controlling for demand and supply side features, any
price differential between paintings of different genres should
disappear.
A similar hypothesis has been largely investigated in the
literature on interindustry wage differentials. The first strand of
studies on this topic (Dickens and Katz 1987; Gibbons and Katz 1992;
Kruger and Summers 1988; Murphy and Topel 1987) found that substantial
wage differentials across sectors (and firms) exist that cannot be
explained by observable worker or firm characteristics. Accordingly,
standard competitive theories of the labor market could hardly
rationalize this evidence (4): one would need to complement them taking
into account compensating wage differentials or market frictions
associated with efficiency wages (informational frictions) or imperfect
mobility of workers (search frictions). Still, common to all of these
early studies is the lack of appropriate data to control properly for
unobservable characteristics of workers and firms. In other words, the
main empirical problems in explaining wage differentials typically
relied on the difficulty of obtaining detailed matched employer-employee
datasets about a competitive labor market and observing the same worker
employed in multiple sectors and in different firms (without selection
on workers switching jobs).
More recently, the emergence of new datasets linking employers and
employees has made this possible. The seminal work of Abowd, Kramarz,
and Margolis (1999) (5) has readdressed the topic of wage differentials
by using unique longitudinal-matched employer-employee data of French
workers and firms: given the matched nature of their panel, they can
control simultaneously for workers and firms' fixed effects, and
they find that most of the interindustry wage differential is explained
by worker effects. Since then, the availability of new data matching
workers to their employers fostered interest for this topic. (6) The
basic message of this literature is that unobserved worker and firm
characteristics matter a lot for wage determination but, although the
inclusion of individual and employer heterogeneity considerably reduces
interindustry differentials, significant differences in wage levels
across sectors persist.
The empirical analysis follows the spirit of this literature
looking at a unique dataset on painters active in Rome in the
seventeenth century, put together by the art historian Richard Spear and
administered by the Getty Research Institute (see Spear and Sohm 2010).
The dataset provides a lot of information on the observable features of
the commissions, of the paintings, and of the painters. (7) Most of all,
the dataset is a matched employer-employee (patron-painter) dataset and
hence it allows us to analyze the interindustry (genre) price
differential exploring the role of worker (artist) and firm (patron)
heterogeneity in the determination of painters' compensation.
The advantage of analyzing a primary artistic market through a
matched employer-employee dataset is that we can observe workers
(painters) that are constantly switching between sectors (genre of
paintings) and between employers (patrons). Moreover, as differences in
the artistic ability (talent) were painter-specific and quintessentially
unobservable (i.e., nonmeasurable), we fully capture them through
artists' fixed effects. Most importantly, unlike in the standard
literature in labor economics, in which each firm could be classified
into a single industry only, the data allow us to identify sectors and
patrons dis-jointly. Indeed in modern labor markets evidence of
interindustry wage differentials could depend, for example, on firm
heterogeneity as firms cannot switch across industries. On the contrary,
in the obtained data each patron could commission paintings belonging to
different genres (or sectors) and so we are able to control for patron
fixed effects. The main contribution is then related to the specific
features of the data that allow us to estimate the intersectoral
compensation differentials using a matched employer-employee while
controlling simultaneously for both workers' and firms'
heterogeneity. (8)
The main result we obtain is that, after controlling for observable
characteristics of both paintings and artists and for both artists'
and patrons' fixed effects, all the price differentials between
genres disappear or become insignificant, which suggests that the labor
market for painters must have been rather competitive and allocated
artists between genres to the point of equalizing the marginal returns
of the different artistic sectors. This is remarkable since we find one
of the earliest microeconometric examples of a competitive mechanism at
work in a labor market of very high skilled workers producing highly
differentiated goods; previous findings of wage equalization in the
literature were based on data on highly interchangeable blue- and
white-collar workers producing similar goods and services. Moreover, the
results are reinforced by the fact that we analyze a market populated by
a small number of agents on both sides of the market, while the
empirical literature on industry wage differential is generally based on
administrative data with thousands of observations. (9)
While arbitrage appears to hold between genres and also between
geographical destinations, we find some evidence of residual price
differentials at the employer level, which we mainly explain in terms of
incentive mechanisms to induce effort in the production of artistic
quality and compensating wage differentials. Interesting results are
related to the moral hazard problem emerging in contractual relations
for figurative paintings in which effort of painters and final quality
of paintings were not contractable ex ante or verifiable ex post. First,
in the case of religious, mythological, and historical paintings, we
show that patrons and artists adopted a typical solution pointed out in
the literature on principal-agent contracts: prices were made
conditional on measurable features of the paintings which were
positively correlated with effort and quality (Holmstrom and Milgrom
1991), one of which was the number of human figures per square meter in
the composition. Second, we find a price premium for altarpieces
destined to public display as in private chapels (within public
churches), which we interpret as evidence of a signaling mechanism in
the commissions by private patrons of artworks that were visible by the
entire community (Nelson and Zeckhauser 2008; Spence 1973).
We find residual price premia paid by foreign kings and foreign
families, such as the Medici from Florence and the Gonzaga from Mantua.
Following the decomposition of variance in Gruetter and Lalive (2009),
we show evidence of negative covariance between artist and patron
effects, and on this basis, we argue that the most convincing
explanation for this is related to compensating wage differentials
rather than efficiency wages. We emphasize that the best paid artists
(for works of similar objective characteristics and destination)
included famous masters such as Maratta, Pietro da Cortona, Guido Reni,
Caravaggio, Vouet, Lorrain, and Poussin. Finally, the rewards of the
painters were increasing with age, consistently with the importance of
on-the-job experience in the artistic market, possibly due to the
development of experimental innovations (Galenson 2006).
This article is organized as follows. Section II describes product
differentiation in the market for paintings and the economic hypotheses
to be tested. Section III illustrates the dataset. Section IV describes
the empirical strategy and the results, and performs some robustness
checks. Finally, Section V concludes.
II. PRICE EQUALIZATION BETWEEN GENRES OF PAINTINGS
Most paintings during the Renaissance were of figurative subject,
mainly religious or mythological, but also allegorical, literary, or
strictly historical (including battles drawn from a real or invented
context), and we define these as figurative paintings. Since the end of
the sixteenth century, however, the raising demand of private buyers
belonging to the high and middle classes (and the influence of Nordic
art) led to the development of new artistic subjects, usually regarded
as minor genres, such as still lifes, landscapes, and so-called genre
paintings (depicting ordinary daily life). These minor genres flourished
during the Baroque age of the seventeenth century.
On the demand side, the substitutability between genres was limited
because of the different decorative roles that the different genres had
in private homes and churches (e.g., portraits had a different purpose
than landscapes within a private building, and only religious paintings
could be destined to churches). On the supply side, painters were
usually engaged in all genres during their career (think of Caravaggio,
Annibale Carracci, or Salvator Rosa to cite some famous eclectic
artists), but often they favored one or two genres more than the others
in their commissions.
While nowadays we are not used to artistic hierarchies based on
paintings' subject, in the seventeenth century the prestige of the
different artistic genres was clearly ranked. The least worthy subjects
were still lifes, imitating inanimate reality without idealization
(Spear and Sohm 2010, 91). The position of portraits in this hierarchy
was variable, but often at a low level for the alleged absence of
creativity in works aimed at merely copying real human figures. Genre
paintings were equally penalized by the lack of decorum and
idealization, and the typical focus on the worst aspects of life, which
was considered vile by contemporary artists and critics (Spear and Sohm
2010, 94). Landscapes and (even more) battles were more valuable than
these genres because of their idealized depictions. The most dignified
and worthy subjects were those depicting creative compositions of ideal
human figures (Spear and Sohm 2010, 91), like historical paintings, and,
most of all, religious and mythological paintings. There is wide
documentary evidence that such a ranking of preferences (and willingness
to pay on the demand side) was well understood between art critics, art
collectors, artists, and intellectuals. (10) Later in the century, it
was even codified by the art academies. (11)
The basic training in different genres and techniques was more or
less common to most artists. Therefore, as long as the market was
competitive and the alternative artistic genres were open to the entry
of new painters, we can hardly imagine that systematic differences in
prices could persist over time between different genres. Painters of
different talent could perfectly choose to be active in different
sectors, but if a given painter could earn more by switching between
genres, this would happen whenever a painting in another genre could
provide a higher compensation. (12) The only differences consistent with
a competitive mechanism could be motivated by compensating
differentials, which at the time were associated with the social
connections that a certain patron could provide to the painter
(providing new and certain commissions in the future). But beyond
exceptional circumstances, a standard arbitrage argument suggests that,
after controlling for observable characteristics of the paintings such
as size, technique, or support, and unobservable characteristics of
painters and commissioners, any price differential between paintings of
different genres should disappear. (13)
A. Price Determinants: The Supply Side
To test the hypothesis of price equalization between genres, we
need to identify the main supply and demand factors that may have
affected prices in the market for paintings of Baroque Rome. On the
supply side, the price of paintings depends first of all on the talent
and the reputation of the painter, which is obviously painter-specific:
the reservation price of painters was clearly increasing in their innate
ability, for instance because better painters had more outside options.
Moreover, the ability to produce high quality could change with the
experience of each painter: the latter may reflect an age/earning
profile dependent on reputational effects emerging with the activity in
the profession (in a Mincerian tradition) or on actual improvements,
that is artistic innovations perceived and priced by the market, that
require a certain experience (Galenson 2006; Galenson and Weinberg
2000).
Other sources of price differential that derive from the supply
side are painting-specific. The most important and measurable feature is
the size of paintings, which reflects the cost of production and
captures many things such as the complexity of the conception as well as
the need of assistants, possibly increasing less than proportionally
with dimension because of scale economies. With regard to figurative
paintings, also the number of figures depicted could have been
increasing the cost of production and therefore the price; however, once
controlling for size, it is not clear if this was still true. Another
factor is the originality of the work: even if the concept of autography
was quite different from our modern understanding (plagiarism was not a
legal issue), copies, often performed by minor painters, implied less
work because they did not require a preliminary creative activity.
The technique adopted could substantially affect costs of
production and prices: compared with oil paintings, frescoes required a
complex preparation (think of ceilings and cupolas, or even large walls)
but also a rapid execution, which could have an ambiguous impact on
costs. (14) Beyond the different costs of production, frescoes (which
were common mainly for figurative paintings and landscapes) provided a
different esthetic perception (and commitment due to the fixed position)
within churches and private residences and, therefore, they could
command different prices. Whether frescoes were paid more or less than
oil paintings remains an open question to be settled at the empirical
level. Finally, in case of oil paintings, the support could also affect
prices: most oil paintings were executed on canvas, but few others of
small size were executed on copper support or other smooth material
(commonly, still lifes and landscapes).
B. Price Determinants: The Demand Side
If we move to analyze the demand side, we expect that different
patrons could have different willingness to pay. The dataset is rich in
information that is useful to identify the patrons (mainly churches or
noble and rich families), the geographical destination of the commission
(Rome, a minor destination in the countryside, a medium city, or a place
outside Italy), the purpose of their commissions (a private collection
and residence, a religious institution, or a private chapel within a
church), and its kind (for a single or multiple paintings). This allows
us also to check for various motivations behind price differentials
generated by different willingness to pay by different patrons.
In the literature on wage differentials at the firm level (i.e.,
employer size-wage effect), there are many explanations on the possible
sources of such differences (e.g., Brown and Medoff 1989; Manning 2003).
The theory of the labor market emphasizing institutions argues that
differences in unionization explain wage differences across firms (e.g.,
Booth 1995). However, unionization was absent in our market. (15)
Equilibrium search theory rationalizes wage differentials across firms
for homogeneous workers due to search frictions (e.g., Postel-Vinay and
Robin 2002; van den Berg and Ridder 1998), which are, however, hard to
apply in a context populated by very high-skilled workers producing
highly differentiated goods. (16) Thus, compensating wage differentials,
in competitive markets, and rent-sharing theories or efficiency wage
considerations, in noncompetitive markets, remain the two most likely
explanations for price differentials.
The first explanation is due to different non-monetary rewards (on
compensating wage differentials, see Hwang, Reed, and Hubbard 1992;
Rosen 1986). If painters had more to gain from commissions from
well-connected Roman families able to guarantee future commissions or
network with other patrons and less to gain from occasional and riskier
commissions from foreign families, they could be indifferent between
lower prices for the first kind of commissions and the second one. Thus,
it is likely to expect negative covariance between artist and patron
effects. Indeed, better artists looked for safer patrons that paid less
for a typical income effect on the supply side (as in labor markets
where high-wage earners purchase safer jobs). (17)
Other sources of price differentials rely on market frictions due
to imperfect observability of workers' effort and
noncontractability (ex ante) of paintings' quality. Theories of
efficiency wages provide a variety of explanations for differential
compensations. On one side, as the reservation price of painters was
increasing in their ability, patrons could pay more to attract the best
painters (Weiss 1980). On the other side, we may expect that the most
susceptible-to-quality patrons were more prone to provide monetary
incentives to induce effort of the painters, and these were ready to
exert higher effort to avoid the risk of working for ordinary patrons
(Shapiro and Stiglitz 1984) or to be paid as painters of lower quality.
For both motivations, it is then likely to expect positive covariance
between artist and patron effects. This is obvious in models of pure
asymmetric information where higher wages are used to select more
productive workers. More productive workers find it more convenient to
choose high wage-high effort contracts relative to low wage-low effort
contracts, and the first kind of contracts are offered exactly to screen
the best workers and increase their effort.
As we could expect from basic contract theory, a variety of
additional incentive mechanisms could emerge in the artist-patron
relationship in case of figurative paintings, for which contractual
agreements were mostly established ex ante and were subject to moral
hazard. (18) We emphasize one aspect related to signaling and one
related to optimal principal-agent contracts. First, as pointed out by
Nelson and Zeckhauser (2008), commissions of altarpieces and frescoes
for private chapels within public churches were visible to the entire
contemporary audience, and could guarantee high visibility for their
patrons in front of the fellow citizens, of the political and
ecclesiastic power (and even of God) and signal what at the time was
called "magnificence." In this sense, we can test a signaling
mechanism a la Spence (1973) for artistic commissions: commissioners
were ready to invest more in order to obtain higher-quality works when a
mechanism of signaling of "magnificence" was active;
therefore, we expect higher prices for paintings addressed to private
chapels as compared with common religious commissions and especially
with other private commissions that were not destined to public display
(such as private palaces and private collections).
Second, we look at the optimal patron-artist contracts from the
perspective of principal-agent theory. As reputational constraints and
efficiency wages did not perfectly align the incentives of patrons and
artists, there was room for further contractual solutions to the moral
hazard problem. In line with the "informativeness principle"
(see Holmstrom and Milgrom 1991), it is reasonable to believe that the
optimal patron-artist contracts had to be based, explicitly or
implicitly, on any verifiable and measurable feature of the painting
that was correlated with effort and quality. The number of figures could
be taken as a proxy for effort and quality for two main reasons. First
of all, the variety and complexity of the composition of human figures
invented by the artist had a positive, although partial, correlation
with quality, and could be summarized exactly by the number of figures.
For instance, we know that pricing by number of figures became a typical
procedure during the early 600s for leading painters such as Guercino,
Domenichino, and Guido Reni. (19) Second, the same hierarchy of genres
could provide a link between number of figures and perceived quality: a
higher number of human figures was increasing on average the space
destined in the painting to subjects of higher perceived value (the
human figures) and reducing the space available for subjects of lower
quality (background landscapes or decorative still lifes). (20)
Taking all this in consideration, we can reformulate the hypothesis
of price equalization between genres in a more appropriate way. Price
equalization should hold, after controlling for observable and
unobservable characteristics of paintings, artists, and patrons, between
still lifes, genre paintings, landscapes, portraits, and all the
figurative paintings with a small number of human figures, allowing for
increasing price premia when the figurative paintings contained a higher
number of figures. (21)
Finally, we also test whether arbitrage was taking place not only
between genres but also between big cities and the countryside. More
specifically, we expect that prices for paintings destined to big and
rich cities (where wealthy demand was) should not be significantly
different from prices of paintings destined to small towns and villages
in the countryside, otherwise profitable opportunities for provincial
painters would have been unexploited. However, notice that this goes
against the common perception at the time, for which Rome was the best
paying market (Spear and Sohm 2010). (22)
III. DATA
In this section, we provide an accurate description of the dataset
and of the variables we consider as price determinants in the empirical
analysis, and we show some descriptive statistics.
A. Description of the Dataset
The empirical analysis is based on a unique dataset established at
the Getty Research Institute (the Payments to Artists Database,
hereafter PAD) which documents payments directly made to artists for the
primary market in seventeenth-century Rome (Spear and Sohm 2010). (23)
The dataset contains approximately a thousand records of payments to
artists and provides a lot of precious information regarding
paintings' and painters' characteristics. To the aim of our
analysis, the most valuable characteristic of the PAD is that it is a
matched employer-employee (patron-painter) dataset. The source of the
information includes original contracts between artists and patrons
(typically for the altarpieces), records of the buyers themselves
(typically for the minor genres, which were not contracted in detail),
evidence from contemporary writers, archives, and inventories. The
survival of all this information for these painters and patrons through
centuries is random, but it is quite reasonable that most of the
information found by art historians actually concerns well-known
painters dealing with famous and wealthy patrons. Therefore, we need to
be aware that this selection excludes from the analysis the lower end of
the art market, which was populated by largely undistinguished craftsmen
supplying low-quality products without any formal agreement with the
buyers. However, such selection allows us to focus on the upper end of
the art market where in general the most famous painters were directly
competing with each other. (24)
In PAD, the information of a typical "arrangement"
between the artist and the patron concerns the record's number, the
artist's name, the title, the subject, the object, the dimension,
the number of figures, the patron's name, the payment date, the
price paid, and the destination (both the original and the current one).
Finally, the records often contain interesting notes providing further
details on the contract, when available. Table 1 describes the structure
of the typical arrangement in PAD.
The original records in the dataset could refer either to a single
or to a multiple commission (i.e., more than one painting for the same
artist-patron contract). When multiple commissions are taken into
account, the total number of single observations increases to about a
thousand and five hundred observations. Whether a painting belongs to a
single or to a multiple commission is generally explicitly indicated in
its title. Knowing the number of paintings for each commission enables
us to control for a possible discount premium. The painting's title
also reports whether each painting represents a copy rather than an
original work.
The painting's price is the dependent variable of the
econometric analysis. The given value is the amount paid to the artist
in silver scudi romani; in some cases, prices were converted from
another currency (such as doubloons, gold scudi, livres tournois,
Spanish real, and pound sterling). (25) In the rare cases in which
prices do not represent an original payment to an artist, they are
identified as evaluation prices or sale prices. In these cases, however,
prices are restricted to the years when the artist was selling his works
because they are probably more representative of the original sale
prices. (26)
In PAD, the "subject" is identified as: sacred,
mythological, allegorical, history, heraldic, literary, battle,
landscape, architectural, portrait, genre, still life, or animals. In
order to obtain variables with an easier interpretation and to link them
with the traditional artistic subject classification, we aggregated more
homogeneous genres among themselves. As a result, we obtained the
following classification: (1) Sacred, (2) Myth (mythological and
allegory), (3) History (history, literary, and heraldic), (4) Battle,
(5) Landscape (landscape and architectural), (6) Portrait, (7) Genre,
and (8) Still life (still life and animal). We defined the first four
groups as belonging to the "figurative" genre, as they
involved traditional compositions of idealized human figures, but we
will also report the distinction in the four subgroups in the
regressions. The number of figures is also given for these figurative
paintings: the full-figure equivalent is reported as a specific number
only when the number of figures is lower than five, whereas full-figure
equivalents are more generally designated "5-10" when they
vary between five and ten and "crowd" when they are greater
than ten or impossible to count. (27)
In the dataset, the "object" refers to both the technique
and the support used by the artist. The former includes drawings,
etching, fresco, mosaic, oil, tapestry, tempera, watercolor; and the
latter includes canvas, panel, mirror, copper, lapis, slate, stucco, and
touchstone. The object also indicates whether a painting was
"Easel," which is used to designate what might also be called
a gallery picture and which could be taken as an indirect information
for its small size when effective size is missing in the dataset. To
guarantee a basic homogeneity in the objects under investigation, we
dropped the observations when the object referred to drawings, mosaics,
tapestries, and watercolors. We do have a few observations for oil
paintings with a support different from canvas and characterized by a
smooth and compact surface (mostly copper, but also mirror, lapis, wood
panel, and others). As a control for the paintings's features, we
included dummies for oil paintings not on canvas and for frescoes. The
dimension has been converted in square meters. In some cases, the
available information is only about one side of the paintings and some
other times only the information "small," "medium,"
"large" or only the size of the frame are given. In all these
circumstances, an appropriate estimate was made (sometimes considering
the distribution of size of comparable subjects in the sample).
The artist's name could either be the name of a single artist
or of more artists, that we considered as copainters. The artist's
name allows us to control for his talent (by including artist fixed
effects) and also to discriminate among painters according to their
origin (local or immigrant). (28) Moreover, as the payment date is
reported in the dataset, by looking at the painters' biographies we
derived the age of the painters at the time in which the artwork was
made.
As demand factors are concerned, in most cases we have information
on the patron's name, which indicates the person or the institution
that commissioned the painting. Patrons could be churches, other
religious institutions (e.g., confraternity or religious orders), the
Vatican church of St. Peter or private patrons. These latter were
generally noble families residing in Rome, as in the case of the
Barberini or the Orsini family, or in other Italian towns, as for the
Gonzaga family in Mantua or the Medici family in Florence. Sometimes
even the Popes active during the century directly commissioned
paintings. Demand for paintings originated also from foreign patrons,
both nobles and kings such as Charles I of England or Louis XIV of
France. Finally, paintings were occasionally bought by rich dealers, as
Mattia Capocaccia, one of the rare traders in paintings, and Fabrizio
Valguamera, apparently trader of (stolen) jewels. (29)
The "destination" in the dataset indicates both the city
and the specific location the artwork was addressed to. The majority of
paintings was commissioned for the city of Rome. However, the dataset
contains many paintings for other important Italian towns, such as
Bologna, Florence, Mantua, or Naples, for minor provincial centers, such
as Caprarola or Frascati, and also for foreign European destinations,
such as London, Madrid, or Paris. The variable destination allows us
also to distinguish between religious and secular locations. Within
religious locations, we are able to discriminate between the case in
which the painting was placed in a family chapel inside a church or in a
public space within the church. Secular locations can be private palaces
or private collections. Therefore, the demand side can be controlled by
patron fixed effects and by looking both at the city where the painting
was addressed to and at the place where the painting was planned to be
positioned. This last variable is likely to be related to the
willingness to pay of the patrons. Table A1 contains the details of the
definitions and sources of the variables included in the regressions.
B. Descriptive Statistics of the Dataset
The payment date in PAD is recorded between 1576 and 1711. In our
analysis, however, we slightly reduced the available observations,
focusing only on the period from 1600 to 1700, for artistic homogeneity
(this is commonly studied as the Baroque century in art history) and
also for monetary reasons: the real value of the silver scudo is known
to have been remarkably stable during that period (see Spear and Sohm
2010), which allows us to focus on the nominal prices in silver coins
without loss of generality. The following descriptive analysis is based
on the observations remaining after filtering data from missing values
(remaining with 1,133 observations). The distribution by genre is shown
in Table 2. Notice that the sacred subjects make the largest share of
the market. Altogether figurative paintings (i.e. sacred, battle,
historical, and mythological subjects) represent about 60% of the
sample. Around 20% are landscapes, whereas portraits and still fifes
cover 9% of the sample each. The associated positive correlation between
the production of paintings per genre and the average prices per genre
is broadly in line with an adjustment mechanism: in sectors where the
willingness to pay was higher, more paintings were produced to equalize
prices compared with sectors where the willingness to pay was lower.
The average price of paintings is 144 scudi, although prices
exhibit a large variation, ranging between 1 scudo romano for some still
fifes and portraits to the 14,000 scudi of the huge fresco by Gaulli
"Triumph of the Name of Jesus" located in the main Roman
church of the Jesuits. In spite of few observations with prices above
1,000 scudi, 90% of the paintings are priced less than 300 scudi, while
the median value is 48. There are some noticeable differences between
the average prices by genre. The highest values are observed for
figurative paintings with historical and sacred subjects at the top
followed by mythological and allegorical subjects, and by battles at the
bottom. Landscapes follow next, while portrait, genre paintings, and
still lifes are the least priced.
The range of variation by dimension is large, with very small
paintings measuring less than 0.5 nr up to the majestic
"Glorification of the Reign of Pope Urban VIII," a 363
[m.sup.2] ceiling fresco by Pietro da Cortona located in Barberini
Palace in Rome. However, apart from few very large paintings, the
average dimension is slightly more than 5 [m.sup.2] while the median is
just 2 [m.sup.2]. Considering genres, the data show that the average
dimension is between 1 and 2 [m.sup.2] for all nonfigurative paintings.
The average dimension is instead more than 8 [m.sup.2] in the case of
figurative paintings. (30)
Turning to patrons, Popes accounted for about 5% of the whole
demand, while around 8% came from religious institutions (excluding St.
Peter's church that alone covers 3.3% of the sample). The paintings
were demanded mainly for private locations, particularly private
collections (62%). Around one quarter of the paintings in the sample was
instead addressed to churches, in same cases sponsored by private
families for their own chapels inside public churches (6.7%). Demand
originated mainly from Rome, but also from other important Italian towns
(16%) and minor centers in the Italian countryside (8%), with large
price differentials for different destinations; about 6% of the
paintings were exported outside Italy.
IV. EMPIRICAL STRATEGY AND RESULTS
We estimate a semi-linear price equation where the natural
logarithm of price is regressed on a set of dummy variables for genres
and on a set of other explanatory variables. Moreover, the PAD has a
matched nature as it relates artist and patron information. This allows
us to estimate the price equation including both painter and patron
fixed effects and hence to evaluate the extent to which price
heterogeneity is related to unobservable characteristics among painters
(artist effect) or among patrons (patron effect). Indeed, unobservable
ability is in general a crucial factor of wage determination. This is
all the more so in the labor market for artists, where the esthetic
value of the artwork, mainly dependent on painter's talent, is one
of the key determinant of its price.
To include both artists' and patrons' fixed effects, we
removed from the sample all the artists and patrons with a single
observation, reducing the number of observations from 1,133 to 1,061.
The sample then comprises a maximum of 1,061 paintings for 87 artists
and 50 patrons (which reduces in number depending on the explanatory
variables we include due to missing values). We should add that the
panel of artists and patrons (panel over paintings, not over time for
our purposes) is strongly unbalanced as it includes artists with a
minimum of 2 observations and a maximum of 43 observations (Poussin),
and patrons with a minimum of 2 to a maximum of 115 observations (the
Chigi family).
We follow the procedure by Abowd, Kramarz, and Margolis (1999) to
analyze the compensation of the painters. (31) Specifically, we estimate
the (log) price of paintings commissioned to an artist i by a patron j
using ordinary least squares (OLS). We include a set of explanatory
variables (i.e., paintings' and artists' characteristics) and
both artist and patron fixed effects. The price equation we estimate is
the following:
[P.sub.in] - [alpha] + [X.sub.in][beta] + [[theta].sub.i] +
[[psi].su.j((i,n)] + [[chi].sub.k(j(i,n))] + [[epsilon].sub.in]
with E[[[epsilon].sub.in] [??] i, n,j(i, n), [X.sub.in]] = 0, (32)
where pin is the logarithm of the price paid to an artist i for a
painting n, [alpha] is a constant, [X.sub.in] denotes the observable
painting-varying exogenous characteristics of both artists and of
paintings (per artist) with coefficient [beta], [[theta].sub.i], is the
pure artist effect, [[psi].sub.j(i,n)] is the pure patron effect for the
patron j(i, n) that has commissioned the painting n to the artist i,
[[chi].sub.k(j(i,n))] is the effect of the genre k(j(i,n))--which is
related to the painting n that is commissioned to an artist i by a
patron j, and [[epsilon].sub.in] is the statistical residual. (33)
We interpret this price equation as a wage equation in which
[p.sub.in] is the compensation of a worker i for a painting n, which is
regressed on a set of observable characteristics of the painting and of
the workers (experience and origin), on the identity of the individual
and on the identity of the employer. The genre to which each painting
belongs could also be interpreted as the industry to which each artist
belongs. This enables us to interpret mobility of artists across
artistic sectors as interindustry workers' mobility and prices
differentials across genres as interindustry wage (compensation)
differentials. One important difference, however, is that while each
firm could be classified into a single industry, here each patron could
commission paintings belonging to different genres.
A. Main Results
The results are summarized in Table 3. In the first column, we
start by estimating a baseline price equation where the natural
logarithm of price is regressed on dummies for genres to highlight the
unconditional price differential between them. As already shown by
descriptive evidence, a sharp ranking of prices can be detected, with
still fifes (the reference category) at the bottom, followed in
increasing order by portraits and genre paintings, and by landscapes and
battles. Figurative paintings are the best paid artworks, with sacred
and mythological subjects with a large number of human figures at the
top. We also performed pairwise t test on the equality of coefficients
between still fifes, portraits, genre paintings, landscapes, and
figurative paintings, and we found that they were all statistically
different from each other with the only exception of the coefficients of
genre paintings and portraits. This result is in fine with the
traditional hierarchy of genres (and with the anecdotal evidence
according to which the relative position of portraits in this hierarchy
was variable). (34) Overall, differences in genre and number of figures
appear to explain only 45% of price variability.
The descriptive analysis has shown a large variation of
paintings' size by subject. More specifically, figurative paintings
have a higher average dimension than other subject paintings. Related
studies on the value of old master paintings in the rest of Italy show
that dimension was a crucial determinant of prices (see Etro and Pagani
2012). Hence, we first want to observe whether the price differential
between genres persists after controlling for paintings' dimension.
The second column shows that indeed it does, confirming the price
differential between genres per square meter, although the relative
price premium diminishes for figurative paintings with a large number of
figures.
As the following step, in the third column, we control for the full
set of paintings' characteristics. Results show that the price
differential still exists, although it is reduced: paintings with the
same objective characteristics are differently paid depending on the
subject they represent. Prices decrease with the number of commissioned
paintings and in the case of oil paintings that are not painted on
canvases, even though only at the 10% level of significance; the
coefficients for the dummies for copies and frescoes are negative as
well. To control for demand effects, in this specification we also
include the geographical destination for which the painting was demanded
and the location where it was planned to be positioned. Quite
interestingly, and in fine with the common perception at the time,
paintings addressed abroad were the best paid, about 200% more than the
reference destination, which is Rome, whereas paintings commissioned for
the countryside were paid about 50% less. As the location destination is
concerned, we can detect a clear ranking where commissions for private
chapels were by far the best paid, those for churches were
"intermediate," and those for private palaces and private
collections (the reference category) were at the bottom. Finally, in
this specification, we introduce two macroeconomic explanatory
variables. The first is the time trend in paintings' prices, which
emphasizes a negative but extremely small trend. The second is a dummy
for the decade following the plague of 1656, which is meant to capture
the effect of the main aggregate shock that may have affected demand
(but also supply) during the century: the effect is not significant.
Both these controls will be omitted in the following specifications,
because the year of execution comoves with the age of the artists and
because of the irrelevance of the aggregate shock. (35)
We then add in the fourth column the artists' observable
characteristics, including the age of the artists and a dummy for
immigrant painters. The price of paintings increases with artist's
age by around 2% per year, in line with that found by Etro and Pagani
(2012) for Venetian paintings in the same period. (36) Concerning the
provenance of the painters, notice that foreign painters were mainly
specialized in minor genres, which were less paid in absolute terms.
However, when controlling for genres, we do not find any evidence of
price differentials between Italian and immigrant painters. (37)
In the fifth column, we enrich the specification by simply adding
patron fixed effects to control for unobservable heterogeneity on the
demand side. Results show that the price differential between genres
still persists, although it is slightly reduced: each patron did pay on
average much more for a figurative painting compared with another
painting of the same size and characteristics but of a different genre,
exactly because the willingness to pay for different genres was variable
according to the hierarchy of preferences for different genres in the
Baroque age. Notice, however, that after controlling for patron fixed
effects the coefficients of the geographical destinations become
nonsignificant at conventional levels, suggesting the existence of an
arbitrage mechanism between destinations with different demand--as
already pointed out in Etro and Pagani (2012) for the Venetian market.
Including only artist fixed effects, in the sixth column, strongly
reduces the price differentials between genres and those coefficients
became almost all insignificant, with the exception of all the
coefficients on figurative paintings. We should emphasize that the value
of the (non-adjusted) R-squared is higher in this specification, which
contains only artist fixed effects (i.e., 0.79) with respect to the
previous one where we included only patron fixed effects (0.76).
Estimation results of the full model of the price equation are
presented in the last column of Table 3. This specification includes
both patrons' and artists' fixed effect (i.e., their
unobservable characteristics to control for individual talent). The
price differential between still lifes and all the other genres finally
disappears: ceteris paribus, a still life was not paid differently from
a portrait, a genre painting, a landscape, or a figurative painting
(with less than five figures) by the same painter. Moreover, notice that
also any price differential between figurative paintings (namely of
historical, mythological, and sacred subject) with a similar number of
figures disappears. This is true for all the three groups of figurative
paintings, that is with less than five figures, with five to ten figures
and with a crowd or more than ten figures (which includes also all the
battles). Therefore, the main hypothesis is fully confirmed: the
artistic sector choice of painters led to the equalization of prices. In
a sense, if figurative paintings were paid more in absolute terms, it
was mainly because better painters were engaged on average in figurative
paintings, and these better painters were paid more also when they were
engaged in other genres. However, minor painters could not switch from
still lifes to figurative paintings to earn extra profits because they
would have been paid less than other painters, exactly enough to make
them indifferent between genres. Which is exactly what a competitive
labor market would have predicted. (38) Notice that a similar result
concerns geographical destinations, whose coefficients are
nonsignificant: arbitrage appears to take place both across sectors
(genres) and between commissions from major destinations and the
countryside.
In Table 4, we report the contributions of artists (fixed effects),
patrons, genres (including figuratives with less than five figures, that
are comparable with the other genres), and other observable
characteristics to explain the variance of logprices (as in Gruetter and
Lalive 2009) after estimation of the full specification in the last
column of Table 3. Pure price differentials between genres provide a
negligible contribution to explain overall price variability (2.52%),
whereas all the observable characteristics together explain around 43%
of it, and artist and patron effects together explain more than 41%,
with a larger role for artists than patrons. These results appear in
line with those of Abowd, Kramarz, and Margolis (1999) and the
subsequent literature, which shows that most of the interindustry wage
differentials can be explained by variation in unobservable
characteristics, mainly of the workers.
At this stage of the analysis, we can comment also on the other
final estimation results. Starting from the effect of paintings'
characteristics, we confirm previous results regarding the positive link
between size and price, with a return above 3% for an additional square
meter beyond the average size; moreover, we find evidence of decreasing
returns given the negative and significant coefficient of the squared
term. The price premium for figurative paintings with a large number of
human figures remains positive and it is statistically significant: plus
150% for figurative paintings with five to ten figures and plus 220% for
those with more than ten figures. (39) This is consistent with the
hypothesis that quantifiable aspects of paintings as the number of
figures depicted were agreed (in preliminary drawings or verbal
communications, as we know it happened for Guercino, Guido Reni, and
others) also with the purpose of insuring a certain effort and quality,
a sort of incentive mechanism to limit moral hazard in line with
standard principal-agent theory (Holmstrom and Milgrom 1991). (40)
The dummy on frescoes is negative and significant: this suggests
that the rapid technique of frescoes, often executed by high-quality
painters, was compensated with a lower payment per square meter.
Nevertheless, in the absence of further information on the time of
execution of frescoes, as compared with oil paintings, we cannot draw
definitive conclusions on the existence of price differentials between
the two techniques. The coefficients for copies and for paintings not on
canvas are negative and significant in specifications not controlling
for artists' characteristics, while in the full specification they
remain negative but they turn insignificant. Therefore, copies (or
replicas of existing paintings) and paintings on copper and other
unusual supports were not paid less, but they were simply carried out by
worse painters: again, there was nothing to gain in specializing in
copies or in paintings on a different support, otherwise a profitable
opportunity would have emerged. Finally, the negative coefficient of the
variable describing the number of paintings in the multiple commissions
highlights the existence of a quantity discount of around 5% for each
additional painting.
On the demand side, interesting results concern the impact of the
place where the artwork was addressed to. As a matter of fact, the same
patron could have different willingness to pay depending on the place
where the painting was planned to be positioned. For instance, patrons
often commissioned paintings for their private chapels within public
churches. Nelson and Zeckhauser (2008) have suggested that higher
investments for these commissions could be motivated to induce higher
quality and signal more magnificence in front of the contemporaries. The
estimates largely confirm this hypothesis: the very high and significant
coefficient for the commissions for private chapels implies that they
were on average better paid than paintings for private collections (the
excluded category) and private palaces (however, the coefficient of
private chapel is higher but not significantly different from that of
other religious destinations).
B. Artist and Patron Effects
Let us consider the role of painters' and patrons'
characteristics in our full specification. The price of paintings
increases with artist's age by around 2% per year, which may be
related to the importance of on-the-job experience in terms of
experimental innovativeness (Galenson 2006; Galenson and Weinberg 2000).
Due to the unbalanced nature of the panel, in Table 3, we report only
the coefficients of the artists with at least eight observations,
because for too few observations it would be difficult to comment on the
values of their coefficients. In order to facilitate interpretation, we
follow the technique used by Kruger and Summers (1988) to compute
industry wage differentials with the method of correction proposed by
Haisken-DeNew and Schmidt (1997) to compute the exact standard errors of
coefficients, and we express results as deviation from an artist-share
weighted mean. The coefficients on the artist dummy variables then show
the artist effect with respect to the sample average. Results show that
some of the most famous painters of the time were the best paid: first
of all the main masters of the typical Baroque art such as Guido Reni,
Pietro da Cortona, and Carlo Maratta (in ascending order), often engaged
in figurative paintings, but also painters who were repeatedly active in
multiple genres such as Caravaggio, who was actually responsible for the
introduction of genre paintings and still lifes in Italian art, or the
French painter Simon Vouet. Among the best paid painters, we also find
two other French painters, Nicolas Poussin and Claude Lorrain, who were
mainly engaged in landscapes.
Let us move to the patrons. As expected, patrons'
characteristics also affect prices. As we did for the artists, in Table
3 we report the coefficients of the patrons with at least eight
observations in the estimation sample, relative to the average patron.
We see that paintings commissioned by the Vatican St. Peter's
church, by foreign kings, and by selected wealthy and powerful families
outside Rome, such as the Medici from Florence and the Gonzaga from
Mantua were paid more than average. Instead, some established families
based in Rome such as the Orsini and the Sacchetti paid less than
average (also Barberini, Borghese, and the Popes have negative, although
not significant at conventional levels, coefficients). Interpreting
these results from a labor market perspective, however, is not
immediate.
As we argued before, one possibility is that the labor market was
not competitive and frictions due to imperfect observability of quality
and effort led to efficiency wage mechanisms aimed at selecting quality
(Weiss 1980) or inducing effort (Shapiro and Stiglitz 1984). These
mechanisms would imply a positive covariance between artist and patron
effects as in this case patrons should pay more to select higher-quality
painters or to induce higher effort. On the contrary, compensating wage
differentials are consistent with a competitive labor market. Working
for some specific patrons provided artists with alternative types of
compensation: examples were the opportunity to access prestigious
networks that provided repeated commissions and so to see an increase in
their future earning profile or at least to reduce their risk in the
process of gaining new commissions. This would generate lower prices for
well-connected Roman families and higher prices for foreign families,
and also a negative covariance between artist and patron effects: better
painters should prefer the safer commissions. (41)
In Table 4, we look at the decomposition of the variance of prices,
as performed by Gruetter and Lalive (2009). The variance-covariance
matrix shows the determinants of the contribution of different effects
in explaining price variability. For instance, most of the contribution
of the artists derives from the variability between artists, but also
from the fact that better paid artists were engaged in commissions with
observable characteristics that were better paid, such as larger
paintings for private chapels. If we look at the contribution of
patrons, we notice a high variability in their payments and again a
positive covariance with the observable characteristics: patrons that
paid more also asked more valuable characteristics, as it should be
intuitive.
More important is to analyze the covariance between patron and
artist effects. Specifically, we find evidence of a negative covariance
of artists and patron effects, which is consistent with the hypothesis
of compensating wage differentials but not with efficiency wages.
Moreover, a closer look at the difference between the single
coefficients of Roman-based patrons with respect to non-Roman ones
provide us with further evidence in favor of the hypothesis of
compensating wage differentials consistently with a competitive labor
market. (42)
C. Robustness Checks
We performed a number of robustness checks on the basic empirical
analysis; the main two will be discussed in this section. (43) Ideally,
one would have liked to observe a dataset in which all patrons demand
paintings of all genres to each artist, which is clearly impossible
within a limited historical sample as ours. However, it is interesting
to check whether price equalization holds where the artists or the
patrons are indeed diversifying their supply or their demand. In the
first column of Table 5, we present the results of a regression in which
we limited the data to the paintings of the "switchers," that
is, the painters that in the dataset have paintings in more than one
genre. The main assumption underlying our result is that the painters
could switch their activity between different genres to exploit any
profitable opportunities, which implies that such opportunities should
disappear in equilibrium: an immediate implication of this is that price
equalization should strictly hold for all the painters that do diversify
within the dataset. The results presented in Table 5 on the switchers
reinforce the main conclusion regarding the validity of the hypothesis
of price equalization between genres. (44)
In the second column of Table 5, we present the results of a
regression in which we limited the data to observations referring to
patrons that demanded paintings of different genres. After this
selection, the number of observations is largely reduced, but the main
results are confirmed. This check can provide additional insights on the
competition mechanism at work: we find again that there are no price
differentials between genres, suggesting the existence of competition
between artists also within commissions from specific patrons.
Finally, we estimate the model including match-specific fixed
effects to control for unobserved characteristics of the artist-patron
match, such as relative bargaining power. Woodcock (2011) presents
direct evidence of the importance of matching in wage determination in
modern labor markets. Fie finds that match effects, which have been
ignored in previous works, are an important determinant of earnings
dispersion, and concludes that specifications that omit match effects
substantially overestimate the returns to experience, attribute too much
variation to personal heterogeneity, and underestimate the extent to
which good workers sort into employment at good firms. In the dataset,
we can identify match effects only for observations referring to artists
and patrons with more than a single transaction and with relationships
with, respectively, more than one patron and more than one artist,
because otherwise the match fixed effects could not be identified. We
are able to identify 64 matches between a corresponding number of
artists and patrons, amounting to 249 observations. Although we cannot
include match effects for all the sample's observations, we
replicate estimates inserting match-specific fixed effects in the
equation to be estimated; results are largely unchanged and they confirm
the main conclusion regarding price equalization (see column 3 of Table
5).
V. CONCLUSIONS
We have analyzed the labor market in the Baroque Roman art sector
using a unique matched painter-patron panel dataset on commissions for
still lifes, landscapes, portraits, genre paintings, battles, and
sacred, mythological, and historical paintings. In line with the
traditional hierarchy of genres, the price differential between them was
high and significant. Adopting a labor economics perspective, we are
able to analyze the intergenre compensation differentials and,
differently from a standard matched employer-employee approach, we
control for both the role of individual and employer heterogeneity in
the determination of artists' compensation.
We find that most of the intergenre price differential is explained
by the variation in average individual heterogeneity across sectors
(genres). This suggests that the labor market for painters was rather
competitive and allocated artists between artistic genres to the point
of equalizing the marginal return of different genres. For each painter,
every commission from a patron for a still life, a portrait, a genre
painting, a landscape, or a comparable figurative painting was equally
profitable at the margin. This reflected an efficient equilibrium of
occupational choice and, incidentally, made it possible for the new
artistic genres to develop and flourish in this and the following
centuries. While we find evidence of arbitrage between genres and also
between geographical destinations, we show some evidence of residual
price differentials at the employer level, which we mainly explain in
terips of incentive mechanisms to induce effort in the production of
artistic quality and compensating wage differentials in competitive
markets.
Additional data on other paintings, patrons, and artists may allow
one to improve the main empirical analysis, to check how the artist
fixed effects evolve with the sales price of previous works and with the
sponsorship of nobles, to check for the relevance of network relations
between artists and patrons and may also allow one to make further
progress in the identification of demand and supply of art. Finally, in
a Schumpeterian perspective, it would be interesting to investigate
whether different compensations affected artistic innovations in certain
locations or certain periods of art history: it was not by chance that
artistic innovation flourished first in the wealthiest cities
characterized by more developed free market economies, as was Rome
during the seventeenth century.
ABBREVIATIONS
OLS: Ordinary Least Squares
PAD: Payments to Artists Database
doi: 10.1111/ecin.12115
Online Early publication June 25, 2014
APPENDIX
TABLE A1
Variables Definitions
Variables Definition
Variables of Interest:
Genre and Number of Figures
Still lifes Dummy = 1 for a still life
Landscapes Dummy = 1 for a landscape
Genre paintings Dummy = 1 for a genre
Portraits Dummy = 1 for a portrait
Figurative paintings (1-4 fig) Dummy = 1 for a "figurative
painting'' (i.e., Sacred or
History or Myth) with 1 to 4 fig.
Sacred (1-4 figures) Interaction Sacred * Figurative (1
to 4 figures)
Historical & Literary (1-4 fig) Interaction History * Figurative
(1 to 4 figures)
Mythological & Allegory (1-4 Interaction Myth * Figurative (1
fig) to 4 figures)
Figurative paintings (5-10 fig) Dummy = 1 for a "figurative
painting" (i.e., Sacred/History/
Myth) with 5 to 10 fig.
Sacred (5-10 fig) Interaction Sacred * Figurative (5
to 10 figures)
Historical & Literary (5-10 fig) Interaction History * Figurative
(5 to 10 figures)
Mythological & Allegory (5-10 Interaction Myth * Figurative (5
fig) to 10 figures)
Figurative paintings (>10 fig) Dummy = 1 for a "figurative
painting" (i.e., Sacred/History/
Myth) with > 10 fig.
Sacred (>10 fig) Interaction Sacred * Figurative
(>10 figures)
Historical & Literary (> 10 fig) Interaction History * Figurative
(>10 figures)
Mythological & Allegory (> 10 Interaction Myth * Figurative (>
fig) 10 figures)
Battles Interaction Battle* Figurative (>
10 figures)
Control Variables:
Paintings' characteristics
Size (square meters) Size in square meters
Size (square meters) squared Size in square meters squared
Number of commissioned Number of paintings for single and
paintings multiple commission
Copy from original Dummy = 1 for a copy form original
Not on canvas Dummy = 1 for paintings on a
support different form canvas
(e.g., copper, etc.)
Fresco Dummy = 1 for frescoes
City destinations
Rome Dummy = 1 for destination to Rome
Minor destinations Dummy = 1 for destination to minor
Italian town (see below)
Medium destinations Dummy = 1 for destination to major
Italian town (see below)
Exports Dummy = 1 for foreign destination
(see below)
Location destinations
Private palace Dummy = 1 for destination to
private palace
Private collection Dummy = 1 for destination to
private collection
Private chapel Dummy = 1 for destination to
private palace
Church and other religious Dummy = 1 for destination to
buildings church and other rel. buildings
Patrons' fixed effects
Private patrons Dummy = 1 when patrons are private
families with at least 2
observations in the sample (see
below)
Churches Dummy = 1 when the patrons are
urban churches
City of Rome Dummy = 1 when the patron is the
City of Rome
Foreign nobles Dummy = 1 when the patron is a
foreign noble
Kings' commission Dummy = 1 when the patron is a
king
Other religious institutions Dummy = 1 for other religious
commissions
Pope's commission Dummy = 1 when the patron is a
Pope
Vatican Dummy = 1 when the patron is the
Vatican
Vatican St. Peters Dummy = 1 when the patron is the
Vatican St. Peter
Artists' characteristics and
fixed effects
Artists Dummy = 1 for artists with at least
2 observations (see below)
Age of artist Difference between payment date
and year of birth
Immigrant Dummy = 1 for artists coming from
outside Italy
Other
Time trend Payment date
Plague Dummy = 1 for the period 1656 1665
(aggregate demand shocks)
Artists Private Patrons Destinations
Abbatini Manfredi Aldobrandini Minor:
Alberti Maratti A1 temps Ariccia
Arpino Maratti & Onofri i Altieri Bagnaia
Baderni Mattia Barberini Bassano di Sutri
Baglione Mei Borghese Bassano Romano
Belloni Miel Bornia Caprarola
Bonzi Mola Borromeo Castel Gandolfo
Both Morandi Brancallero Castel San Pietro
Brandi Nuzzi Campello Catania
Bril Pace Capocaccia Cesena
Brueghel Passerotti Cardelli Fano
Camassei Pellegrini Cerasi Fara
Caravaggio Perfetta Chigi Foligno
Caroselli Pomarancio Colonna Frascati
Carracci Porpora Correggio Lanuvio
Cerquozzi Poussin Corsini San Quirico d'Orcia
Cerrini G. Preti de Rossi Spoleto
Chiari M. Preti Farnese Tivoli
Ciampelli Reni Filomarino
Cigoli G. Romanelli Furgotto Medium:
Codazzi U. Romanelli Giustiniani Arezzo
Courtois Rondoni Gonzaga Ferrara
Cozza Rosa Guicciardini Florence
Cresti Sacchi Mattei Mantua
Cortona Salini Mazarin Messina
V. de Boulogne Saracen i Medici Milan
Domenichino Sassoferrato Orsini Naples
Dughet G. Stanchi Pamphilj Palermo
Elsheimer N. Stanchi Peretti Montalto Perugia
Ferri Stella Pointel Pistoia
F. Napoletano Stom Riviera Siena
Galli Swanevelt Roscioli Venice
Gaulli Tanari Rospigliosi
A. Gentileschi Tassi Ruffo Exports:
G. Gimignani Tempesta Sacchetti Antwerp
L. Gimignani Tomasini Santacroce Austria
Gramatica Trevisani Santori Dalmazia
Grimaldi Turchi Savoia London
Honthorst Vajani Sfondrato Madrid
Jannetti F. Vanni Spada Paris
Laer R. Vanni Valguarnera Rouen
Lanfranco Viola Switzerland
Leoni Vouet
Lorrain
TABLE A2
Artist Specialization by Genre
Number of Genres
1 2 3 Total
No. of artist 50 20 9 79
of which:
Still life 9 1 2 12
Genre 3 1 4 8
Landscape 9 3 6 18
Portrait 0 10 4 14
Figurative 29 19 8 56
% observations 52 29 19 100
REFERENCES
Abowd. J. M.. F. Kramarz, and D. N. Margolis. "High Wage
Workers and High Wage Firms." Econometrica, 67(2), 1999. 251-333.
Abowd, J. M., F. Kramarz, D. N. Margolis, and K. R. Troske.
"The Relative Importance of Employer and Employee Effects on
Compensation: A Comparison of France and the United States."
Journal of the Japanese and International Economies, 15, 2001, 419-36.
Abowd, J. M., F. Kramarz, P. Lengermann, K. L. McKinney, and S.
Roux. "Persistent Inter-industry Wage Differences: Rent Sharing and
Opportunity Costs." IZA Journal of Labor Economics, 1(7), 2012.
Abowd, J., F. Kramarz, and S. Woodcock. "Econometric Analyses
of Linked Employer-Employee Data," in The Econometrics of Panel
Data, edited by L. Matyas and P. Sevestre. Berlin: Springer Netherlands,
2008, 727-60.
Ackerberg, D. A., and M. Botticini. "Endogenous Matching and
the Empirical Determinants of Contract Form." Journal of Political
Economy, 110(3), 2002, 564-92.
Ashenfelter, O., and K. Graddy. "Auctions and the Price of
Art." Journal of Economic Literature, 41. 2003, 763-86.
Bertrand, M., C. Goldin, and L. F. Katz. "Dynamics of the
Gender Gap for Young Professionals in the Corporate and Financial
Sectors." American Economic Journal: Applied Economics, 2, 2010,
228-52.
Booth, A. The Economics of Trade Unions. Cambridge: Cambridge
University Press, 1995.
Brown, C., and J. Medoff. "The Employer Size-Wage
Effect." Journal of Political Economy, 97(5), 1989, 1027-59.
Card, D., J. Heining, and P. Kline. "Workplace Heterogeneity
and the Rise of West German Wage Inequality." NBER Working Paper
No. 18522, 2012.
De Marchi, N., and H. Van Miegroet, ed. Mapping Markets for
Paintings in Europe 1450-1750. Turnhout, Belgium: Brepols, 2006.
Dickens, W. T., and L. F. Katz. "Inter-Industry Wage
Differences and Industry Characteristics," in Unemployment and the
Structure of Labor Markets, edited by K. Lang and J. S. Leonard. Oxford:
Basil Blackwell, 1987.
Eeckhout, J., and P. Kircher. "Identifying Sorting-In
Theory." Review of Economic Studies, 78(3). 2011, 872-906.
Enggass, R., and J. Brown. Italian and Spanish Art, 16001750:
Sources and Documents. Evanston, IL: Northwestern University Press,
1970.
Etro, F. The Economics of Renaissance. Mimeo, Ca' Foscari
University, 2013.
Etro, F., and L. Pagani. "The Market for Paintings in Italy
during the Seventeenth Century." Journal of Economic History,
72(2), 2012, 423-47.
--. "The Market for Paintings in the Venetian Republic from
Renaissance to Rococo." Journal of Cultural Economics, 37(4), 2013,
391-415.
Etro, F., and E. Stepanova. "The Market for Paintings in the
Netherlands during the Seventeenth Century." Working Paper No. 16,
University of Venice, Ca' Foscari, Department of Economics, 2013.
Galenson, D. Old Masters and Young Geniuses: The Two Life Cycles of
Artistic Creativity. Princeton, NJ: Princeton University Press, 2006.
Galenson, D., and B. A. Weinberg. "Age and the Quality of
Work: The Case of Modern American Painters." Journal of Political
Economy, 108(4), 2000, 761-77.
Gibbons, R., and L. Katz. "Does Unmeasured Ability Explain
Inter-Industry Wage Differentials?" Review of Economic Studies, 59,
1992, 515-35.
Goux, D., and E. Maurin. "Persistence of Interindustry Wage
Differentials: A Reexamination Using Matched Worker-Firm Panel
Data." Journal of Labor Economics, 17(3), 1999, 492-533.
Graddy, K. "Taste Endures! The Rankings of Roger de Piles
([dagger] 1709) and Three Centures of Art Prices." Journal of
Economic History, 73(3), 2013, 766-91.
Gruetter, M., and R. Lalive. "The Importance of Firms in Wage
Determination." Labour Economics, 16(2), 2009, 149-60.
Haisken-De New. J. P., and C. M. Schmidt. "Interindustry and
Interregion Differentials: Mechanics and Interpretation." Review of
Economics and Statistics, 79, 1997, 516-21.
Haskell. F. Patrons and Painters: Art and Society in Baroque Italy.
New York: Alfred N. Knopf, 1982.
Hellmanzik, C. "Location Matters: Estimating Cluster Premiums
for Prominent Modern Artists." European Economic Review, 54, 2010,
199-218.
Holmstrom. B., and P. Milgrom. "Multitask Principal-Agent
Analyses: Incentive Contracts, Asset Ownership, and lob Design."
Journal of Law, Economics, and Organization, 1, 1991, 24-52.
Hwang, H.-S., W. R. Reed, and C. Hubbard. "Compensating Wage
Differentials and Unobserved Productivity." Journal of Political
Economy, 100(4), 1992, 835-58.
Kruger, A. B., and L. H. Summers. "Efficiency Wages and the
Inter-Industry Wage Structure." Econometrica, 56(2), 1988, 259-93.
Manning, A. Monopsony in Motion: Imperfect Competition in Labor
Markets. Princeton, NJ: Princeton University Press, 2003.
Montias, J. Art at Auction in 17th Century Amsterdam. Amsterdam:
Amsterdam University Press, 2002.
Murphy, K. M., and R. H. Topel. "Unemployment, Risk, and
Earnings: Testing for Equalizing Wage Differences in the Labor
Markets," in Unemployment and the Structure of Labor Markets,
edited by K. Lang and J. S. Leonard. Oxford: Basil Blackwell, 1987.
Nelson, J. K., and R. J. Zeckhauser. The Patron's Payoff.
Conspicuous Commission in Italian Renaissance Art. Princeton, NJ:
Princeton University Press, 2008.
O'Malley, M. The Business of Art. Contracts and the
Commissioning Process in Renaissance Italy. New Haven, CT: Yale
University Press, 2005.
Postel-Vinay, F., and J.-M. Robin. "Equilibrium Wage
Dispersion with Worker and Employer Heterogeneity." Econometrica,
70(6). 2002, 2295-350.
Rosen, S. "The Theory of Equalizing Differences," in
Handbook of Labor Economics, vol. 1, edited by O. C. Ashenfelter and R.
Layard. Amsterdam: North Holland, 1986, 641-92.
Shapiro, C., and J. Stiglitz. "Equilibrium Unemployment as a
Worker Discipline Device." American Economic Review, 74, 1984,
433-44.
Spear, R., and P. Sohm. Painting for Profit. The Economic Lives of
Seventeenth-Century Italian Painters. New Haven, CT; London: Yale
University Press, 2010.
Spence, M. "Job Market Signaling." Quarterly Journal of
Economics, 87(3), 1973, 355-74.
van den Berg, G. J., and G. Ridder. "An Empirical Equilibrium
Search Model of the Labor Market." Econometrica, 66(5), 1998,
1183-221.
Weiss, A. M. "Job Queues and Layoffs in Labor Markets with
Flexible Wages." Journal of Political Economy, 88(3), 1980, 526-38.
Woodcock, S. "Wage Differentials in the Presence of Unobserved
Worker, Firm, and Match Heterogeneity." Labour Economics, 15(4),
2008, 771-93.
--. Match Effect. Mimeo, Simon Fraser University, 2011.
(1.) For analysis of the secondary market in historical periods see
Montias (2002) and Etro and Stepanova (2013), who focus on the Dutch
auctions of the seventeenth century.
(2.) Etro (2013) analyzes data from the Italian Renaissance market.
Etro and Pagani (2012, 2013) have mainly analyzed the Venetian market of
altarpieces. For a classic work on Baroque patrons see Haskell (1982).
(3.) See Spear and Sohm (2010, 91).
(4.) Which is true also in the case of highly skilled workers such
as the painters of the dataset (e.g., see Bertrand et al. 2010).
(5.) See also Goux and Maurin (1999) and more recently Gruetter and
Lalive (2009) and Woodcock (2008) for related analysis.
(6.) For instance, after decomposing wage rates into a worker
effect and a firm effect, Abowd et al. (2001) and Abowd et al. (2012)
compute the share of interindustry wage differences attributable to
worker and to firm effect in France and the United States. The second
work relates the part arising from individual heterogeneity to the
worker's opportunity wage rate and the part arising from employer
heterogeneity to product market quasi-rents and relative bargaining
power. Moreover, different modeling techniques have been developed to
address the econometric challenges related to longitudinal matched
employer-employee data (see the survey in Abowd, Kramarz, and Woodcock
2008). Recently, Card, Heining, and Kline (2012) have used the
methodology to investigate the evolution of wage inequality.
(7.) We know the age at which painters executed the works: since
they received a similar general training before becoming independent
masters, but kept training and improving their ability and reputation
with time, on-the-job experience is the main observable characteristic
of the painters that we use.
(8.) In labor terms, sector or industry is a characteristic of the
firm, thus the definition of the pure industry effect (for example in
Abowd, Kramarz, and Margolis 1999) is simply the correct aggregation of
the pure firm effect within the industry. This is not the case in our
context.
(9.) Abowd, Kramarz, and Margolis (1999) study a longitudinal
sample of the French labor market with over one million workers from
more than five-hundred thousand employing firms.
(10.) Vincenzo Giustiniani (1564-1637), a famous art collector, in
a letter distinguished 12 "categories, concerning both the methods
of painting and the rankings of painters" with a clear hierarchy
(see Enggass and Brown 1970). The worst three categories, or "
methods" concerned copies. At a higher level of Giustinani's
ranking were portraits and still lifes. At an even higher level, he
placed different kinds of landscape paintings. The best categories were
about figurative paintings, including battles and, a step above,
historical subjects divided into subcategories differentiated only from
a stylistic point of view. Similar views were expressed by other experts
in the following decades.
(11.) The hierarchy of genres became a source of intellectual
debate in the European art academies. A shared view was later codified
in a famous lecture given by the art critic Andre Felibien at the French
Academy (Conferences de l'Academic Royale de Peinture et de
Sculpture pendant l'annee 1667). His influential hierarchy of the
genres, ranked still lifes in the lowest position and figurative
paintings at the highest level. For an economic analysis of ancient art
critique, see Graddy (2013).
(12.) In a Schumpeterian perspective, the same innovation of new
genres in the Baroque period can be justified by the new economic
compensation that minor genres started to deliver to the painters. See
Etro and Pagani (2013) on Schumpeterian patterns in the Venetian art
market.
(13.) Notice that we are not claiming that each artist was equally
good in each genre, because clearly there could be different talents and
skills in different genres (and specialization could induce improvements
in a single genre and not in the others). The hypothesis is that if
artists could switch between genres, they would do it whenever painting
a different genre could provide a higher payment. Therefore, any
equilibrium would allocate works (and workers) between genres to the
point of equalizing prices.
(14.) As Spear and Sohm (2010) notice, "more data are required
before it can be said if it definitely was cheaper to paint in fresco
than oil, not only because there were so many variables in the quality
and quantities of pigments used in different jobs, but also because a
fresco painter might or might not have been responsible for the cost of
preparing a wall, or an oil painter for buying his canvases and
stretcher. I suspect that generally fresco was the cheaper medium by
measure, but even so that does not take into account the overhead of
hiring more assistants for elaborate projects" (56-57).
(15.) More specifically, the painters did belong to a guild, but
this was not able to affect prices in any effective way (Spear and Sohm
2010).
(16.) Bargaining power on the product market and the rents stemming
from it cannot be a source of price differentials at the patron level,
because in our context the workers (artists) produce a consumption good
for the final consumers and they are not hired as inputs in the
production of goods to be sold on another market.
(17.) See the discussion in Gruetter and Lalive (2009).
(18.) For an interesting analysis of contractual relations in
premodern time, see Ackerberg and Botticini (2002).
(19.) Guercino claimed to commit to a fixed price of 100 scudi per
full-length figure (50 for half-length figure, 25 for heads); however,
this could be part of a sophisticated bargaining technique because
deviations from this "commitment" were the rule rather than
the exception. In a letter of 1628, Guido Reni argued that the low-level
painters could not obtain more than 2 or 3 scudi for large life-size
figures and ordinary painters could ask at most 15 scudi per figure,
while an extraordinary painter like himself could name his own price on
the basis of the quality of his work independently from size and number
of figures (Spear and Sohm 2010).
(20.) Moreover, painters were often focusing their own effort on
human figures and especially on difficult parts as the heads, delegating
less relevant parts (including background decorations, landscapes, and
still lifes) to their own assistants. Accordingly, a higher number of
figures was a proxy for a wider direct intervention of the master
painters in the overall execution, and consequently for higher quality.
(21.) To be as conservative as possible, we define figurative
paintings with a small number of figures as those with at most four
human figures. This is particularly conservative for our purposes
because portraits depicted almost always a single human figure.
(22.) In 1625, Fra Atanasio, an art dealer who was negotiating an
altarpiece by Giovanni Battista Crespi called Cerano in Milan, told the
patron that the painter would have probably accepted 250 scudi, but also
that if Cerano were to go to Rome he would be paid double because, he
added, Rome is "where you go to get rich" (Spear and Sohm
2010, 233).
(23.) The original dataset is available online at:
http://www.getty.edu/research/tools/provenance/payments_to_artists/index.html.
(24.) The dataset presents quite a few missing values, which in
fact markedly reduced the total number of available observations. More
specifically, we decided to delete all the records containing
simultaneous missing information on the subject (or genre), title, and
dimension. However, in order to obtain a number of observations as large
as possible we decided to make a few guesses regarding the missing
values of a given variable, provided that all the other crucial
information was certain. All these guesses are described more precisely
in a Data Appendix available on request.
(25.) Notice that some payments were made in kind (with wine,
wheat, cheese, diamonds, even flowers, and marzipan): however, their
equivalent cash value is cited in the documents and reported in PAD.
(26.) As all artists in PAD (with the notable exception of
Artemisia Gentileschi) are male, we will use the male pronoun throughout
all this work.
(27.) In PAD, data were assembled with the method for figure
counting used by the Deputies of the Cappella del Tesoro di San Gennaro
in Naples in the 1630s. Other than counting what obviously were full or
half figures, they counted a certain number of putti as the equivalent
of a full figure.
(28.) Rome attracted many foreign painters, mainly Dutch and
Flemish (e.g.. Both, Bril, Brueghel, and Honthorst present in the
dataset) or French (such as Lorrain, Dughet, and Poussin), who were
often focused on the minor genres: mainly still lifes and genre
paintings for the Dutch and the Flemish, and landscapes for the French.
(29.) Only in the case of Capocaccia, eight prices derive from a
resale (to the Cardinal Francesco Maidalchini); therefore, they should
be interpreted as an (upperbound to the) estimate of the original
prices.
(30.) When considering figurative paintings, an important attribute
to consider is the number of figures depicted. The largest share of
paintings (almost 50%) have a low number of figures (from 1 to 4),
around one-third have an intermediate value (between 5 and 10) while
only 17% of figurative paintings contain more than 10 figures.
(31.) Notice that Eeckhout and Kircher (2011) have recently
criticized the procedure by Abowd, Kramarz, and Margolis (1999) to
measure sorting between employers and workers, showing that in general
matching models the direction of sorting depends on the cross derivative
of the production function with respect to both firms and workers, whose
sign is ambiguous. In our market, the production function (of quality of
the paintings) depends only on the workers (the artists) and not on the
employers (the patrons); therefore, this theoretical critique does not
apply.
(32.) This assumption is usually problematic in modern labor
markets, because it requires exogenous mobility between jobs. When
voluntary, a transition to a new job is rare and aimed at increasing the
wage, and therefore it can be systematically biased by endogenous
factors. This is probably not the case in our market, where painters
were constantly switching from a job (and a patron) to another.
(33.) After obtaining the least square parameter estimates
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] we can determine the
importance of different factors in price determination with the
procedure of Gruetter and Lalive (2009). Using the following
decomposition of the variance of the logarithm of the price:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
we can express the importance of each factor as the ratio between
the covariance of its contribution with the price and the overall
variance. In particular, the importance of the artists effects is:
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]
and the same for the other components. Notice that cov
[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] implies positive
sorting of artists across patrons effects, and cov ([MATHEMATICAL
EXPRESSION NOT REPRODUCIBLE IN ASCII]) implies negative sorting. Results
are presented in Table 4.
(34.) See Spear and Sohm (2010). Incidentally, this result is also
consistent with the descriptive evidence reported in Table 2, where
slightly increasing the number of observations is enough to invert the
rank between genre and portrait with respect to the one emerging from
the baseline regression. Notice that historical subjects and battles
appear to be less paid than other figurative subjects, at least in
compositions with a large number of figures. Also this is in line with
the hierarchy of genres, because idealized subjects (as religious and
mythological subjects) were better considered than realistic
(historical) ones.
(35.) Controlling for age, the time trend becomes indeed
irrelevant. And the main results remain the same if we replace age with
the year of execution in the full specification. We have also used 10 or
20 years dummy variables to control for time effects and the results are
unchanged.
(36.) We also included age squared among the regressors but its
coefficient is very small and highly insignificant.
(37.) Results will not be displayed in the full specification due
to collinearity with the artists' fixed effects.
(38.) Table A2 shows that the number of actual "
switchers" in the sample is considerable and corresponds to 37% of
all the artists in the estimated sample (and to about half of the number
of observations because they have more paintings in the dataset on
average). Of course, most of the artists may have been engaged in
multiple genres even if this is not reported in the limited sample.
(39.) We remark that the price premium for figurative paintings
with a small number of figures would have not only been statistically
insignificant, but also economically insignificant, had we limited the
category to a lower number of figures (up to two or three). We are
grateful to Deirdre McCloskey for discussion on this point.
(40.) This hypothesis needs further verification. The price premium
for number of figures is confirmed in a dataset with only figurative
paintings. Most important, it disappears for the subset of figurative
paintings whose price was estimated ex post. This may confirm that a
price premium emerges only in ex ante contracts as a partial solution to
the moral hazard problem. We are grateful to Paul Milgrom and Julio
Rotemberg for lively discussions on this point.
(41.) A negative covariance between artist and patron effects could
also be simply explained by different bargaining power in noncompetitive
markets. However, a difference in the (negative) signs of the
well-connected Roman patrons (who were in fact able to guarantee access
to networks and to repeated commissions), with respect to the (positive)
signs of the non-Roman ones, would make compensating wage differentials
a most convincing explanation. A related point is that the observed
cross-patrons price differentials may reflect unobserved characteristics
of the artist-patron match indicating relative bargaining power. In the
next section, we replicate estimation including match-specific fixed
effects.
(42.) Extremely rich and powerful foreign patrons, as foreign
kings, the Medici, and the Gonzaga did have to pay a lot to obtain works
from artists active in Rome, but rarely from the very top ones. Instead,
some important families such as Orsini and Sacchetti managed to purchase
repeatedly from top painters at good prices. Remarkably, the only art
dealer in the dataset, Capocaccia, has a negative coefficient: probably
it was for his services to the painters as a distributor of paintings
across Roman families that he managed to obtain good deals from the best
painters. Only the exceptional prices for the decoration of St.
Peter's church do not appear in line with the hypothesis of
compensating wage differentials, but this case is unique under many
dimensions.
(43.) In unreported regressions, we also excluded outliers (in
particular, paintings with extremely high price or size), frescoes and
paintings on a different support than the most common canvas, we
extended the dataset to the period 1576-1711, and we controlled for the
evolution of the artist fixed effects with age and for an alternative
specification of paintings' size. These additional results confirm
the main results and are available on request.
(44.) It is important to remark that we know from art historical
sources that almost all the painters in the dataset did paint
commissions from most genres (although these are not documented in the
dataset).
FEDERICO ETRO, SILVIA MARCHESI and LAURA PAGANI *
* We would like to thank Lorenzo Cappellari and Barbara Petrongolo
for their insightful and constructive comments and suggestions. We also
thank Orley Ashenfelter, Karol Jan Borowiecki, Emilio Colombo, Tiziana
Cuccia, Maria De Paola, Giulio Fella, Davide Fiaschi, James Fleckman,
Francis Kramarz, Emanuela Marrocu, Ulisse Mangialaio, Deirdre McCloskey,
Paul Milgrom, John O'Hagan, Alessia Paccagnini, Mario Padula,
Valeria Pinchera, Julio Rotemberg, Andrei Shleifer, and Richard Spear.
Finally, we thank participants at the XXVIII Italian Conference of Labor
Economics (Rome, 2013), the European Association of Labor Economist
(Torino, 2013), the VI European Workshop on Applied Cultural Economics
(Ljublijana, 2013), the III Workshop on Institutions, Individual
Behavior and Economic Outcomes (Argentiera, 2012), and at the 53 SIE
Annual Meeting (Matera, 2012), and seminar participants at the
Copenhagen Business School, LSE, University of Pisa, Queen Mary
University of London, the University of Milano Bicocca (CISEPS), the
University of Southern Denmark, and the University of Venice for useful
comments.
Etro: Professor, Department of Economics, University of Venice,
Ca' Foscari, 30121 Venezia, Italy. Phone +39 041 234 9172, Fax +39
041 234 9176, E-mail federico.etro@ unive.it
Marchesi: Professor, Department of Economics, Management and
Statistics, University of Milan, Bicocca, 20126 Milano, Italy. Phone +39
02 64483057, Fax +39 02 64483085, E-mail silvia.marchesi@unimib.it
Pagani: Professor, Department of Economics, Management and
Statistics, University of Milan, Bicocca, 20126 Milano, Italy. Phone +39
02 64483060, Fax +39 02 64483085, E-mail laura.pagani@unimib.it
TABLE 1
Typical Record in PAD
PI Record P-268
Artist name CARAVAGGIO,
MICHELANGELO MERISI
DA (Italian)
Title Supper at Emmaus
Subject Sacred
Object type Oil (Easel)
Dimensions 141 x 196 cm
Figures 4 half lengths
Patron name Mattei, Ciriaco
Payment date 1601
Price paid 150 scudi
City (destination) Rome, Mattei collection
Present location London, National Gallery
Note: PI, provenance index.
TABLE 2
Descriptive Statistics of the Original Dataset (1,133 observations)
Genre distribution (%)
Still Portrait Genre Landscape
9 9 3 21
Average price by genre (in scudi romani)
Still Portrait Genre Landscape
17 39 25 66
Average size by genre (in square meters)
Still Portrait Genre Landscape
1.1 1.9 1.4 1.9
Number of figures distribution (%) Medium (5-10)
Low (1-4)
49 33
Object type distribution (%)
Not on canvas support Fresco
6 7
Patrons (%)
Other religious King Foreign noble St. Peter
0.7 1 2 3.3
Location destination (%)
Private chapel Private palace
6.7 12
City destination (%)
Foreign Rome
6 70
Average price by destination (in scudi romani)
Foreign Rome
238 146
Genre distribution (%)
Still Battle Myth Sacred History
9 3 7 45 3
Average price by genre (in scudi romani)
Still Battle Myth Sacred History
17 73 202 242 262
Average size by genre (in square meters)
Still Battle Myth Sacred History
1.1 1.5 15.3 7.3 11.5
Number of figures distribution (%) High (crowd)
Low (1-4)
49 17
Object type distribution (%)
Not on canvas support Copy
6 3
Patrons (%)
Other religious Pope Church Private families
0.7 5 7 81
Location destination (%)
Private chapel Church Private collection
6.7 19 62
City destination (%)
Foreign Medium Minor
6 16 8
Average price by destination (in scudi romani)
Foreign Medium Minor
238 137 127
TABLE 3
Log Price Determination
Baseline
Baseline With Size
(1) (2)
Still life (omitted)
Portrait 0.667 *** 0.627 ***
(0.156) (0.154)
Genre 0.686 *** 0.681 ***
(0.216) (0.211)
Landscape 1.411 *** 1.361 ***
(0.147) (0.146)
Figurative (<5 fig) 1 946 1.809 ***
(0.147) (0.150)
Sacred < 5 (omitted)
History < 5 -0.048 0.037
(0.123) (0.125)
Myth < 5 -0.256 -0.237
(0.168) (0.162)
Figurative (5-10 fig) 2.897 *** 2.535 ***
(0.162) (0.163)
Sacred 5-10 (omitted)
History 5-10 -0.684 *** -0.423 *
(0.232) (0.238)
Myth 5-10 -0.241 -0.039
(0.467) (0.441)
Figurative (>10 fig) 3.671 *** 2.780 ***
(0.219) (0.210)
Sacred > 10 (omitted)
History > 10 -1.651 *** -0.948 ***
(0.375) (0.335)
Myth >10 -0.088 -0.494 *
(0.351) (0.296)
Battle >10 -2.354 *** -1.484 ***
(0.355) (0.340)
Size 0.052 ***
(0.009)
Size^2 -0.0001 ***
(0.000)
# commissions
Copy
Not on canvas
Fresco
Private chapel
Church and other rel
Private palace
Private collection (omitted)
Minor destinations
Medium destinations
Exports
Rome (omitted)
Age of artist
Immigrant
Trend
Plague
Constant 2.161 *** 2.103 ***
(0.118) (0.116)
Artist fixed effects No No
Patron fixed effects No No
Observations 762 726
[R.sup.2] 0.447 0.507
Artists with at Least Eight Observations in the Estimated Sample
Arpino 0.169
(0.291)
Baglione -0.015
(0.350)
Belloni -0.941 ***
(0.330)
Both 0.346
(0.311)
Bril -0.592 *
(0.336)
Camassei -0.052
(0.343)
Caravaggio 0.793 **
(0.336)
Cerquozzi 0.004
(0.431)
Courtois -0.709 **
(0.340)
Cortona 0.861 ***
(0.221)
V. de Boulogne 0.423
(0.315)
Domenichino 0.094
(0.190)
Dughet -0.532
(0.328)
Ferri 0.245
(0.207)
Gaulli 0.212
(0.242)
Gimignani -0.314 *
(0.175)
Grimaldi 0.261
(0.323)
Jannetti -2.052 ***
(0.495)
Lanfranco 0.475 **
(0.215)
Patrons with at Least Eight Observations in the Estimated Sample
Aldobrandini -0.294
(0.314)
Altemps -0.381
(0.540)
Barberini -0.111
(0.118)
Borghese 0.027
(0.236)
Patrons with at Least Eight Observations in the Estimated Sample
Capocaccia -0.840 **
(0.338)
Chigi 0.082
(0.146)
Church 0.067
(0.180)
Colonna 0.033
(0.238)
Filomarino -1.566 ***
(0.395)
Foreign nobles 0.226
(0.226)
Gonzaga 1.678 **
(0.704)
King 1.405
(0.401)
Paintings Artist
Characteristics Characteristics
(3) (4)
Still life (omitted)
Portrait 0.508 *** 0.705 ***
(0.134) (0.137)
Genre 0.521 *** 0.433 *
(0.200) (0.230)
Landscape 1.081 *** 1 073
(0.132) (0.148)
Figurative (<5 fig) 1 315 1 414
(0.148) (0.145)
Sacred < 5 (omitted)
History < 5 0.044 -0.031
(0.220) (0.240)
Myth < 5 0.033 -0.022
(0.192) (0.189)
Figurative (5-10 fig) 1.768 *** 1.814 ***
(0.157) (0.157)
Sacred 5-10 (omitted)
History 5-10 -0.116 -0.013
(0.266) (0.216)
Myth 5-10 0.011 -0.011
(0.393) (0.369)
Figurative (>10 fig) 2.030 *** 2.280 ***
(0.222) (0.218)
Sacred > 10 (omitted)
History >10 -0.185 -0.333
(0.284) (0.317)
Myth >10 0.316 0.080
(0.302) (0.295)
Battle >10 -0.895 *** -1.134 ***
(0.337) (0.334)
Size 0.048 *** 0.044 ***
(0.009) (0.009)
Size^2 -0.0001 *** -0.0001 ***
(0.000) (0.000)
# commissions -0.037 *** -0.049 ***
(0.009) (0.009)
Copy -0.597 *** -0.843 ***
(0.220) (0.265)
Not on canvas -0.282 * -0.311 **
(0.154) (0.149)
Fresco -0.641 *** -0.535 ***
(0.187) (0.197)
Private chapel 1.031 *** 1.186 ***
(0.134) (0.130)
Church and other rel 0.902 *** 0.901 ***
(0.130) (0.127)
Private palace 0.633 *** 0.786 ***
(0.146) (0.151)
Private collection (omitted)
Minor destinations -0.706 *** -0.651 ***
(0.155) (0.152)
Medium destinations 0.505 *** 0.475 ***
(0.108) (0.110)
Exports 1.111 *** 0.801 ***
(0.156) (0.167)
Rome (omitted)
Age of artist 0.021 ***
(0.003)
Immigrant -0.050
(0.105)
Trend -0.006 ***
(0.002)
Plague 0.153
(0.106)
Constant 12.411 *** 1.354 ***
(2.717) (0.184)
Artist fixed effects No No
Patron fixed effects No No
Observations 716 716
[R.sup.2] 0.634 0.652
Artists with at Least Eight Observations in the Estimated Sample
Arpino Leoni
Baglione Lorrain
Belloni Maratti
Both Mei
Bril Morandi
Camassei F. Napoletano
Caravaggio Nuzzi
Cerquozzi Pace
Courtois Poussin
Cortona Mattia Preti
V. de Boulogne Reni
Domenichino G. Romanelli
Dughet Rosa
Ferri G. Sacchi
Gaulli Salini
Gimignani Stanchi
Grimaldi Tempesta
Jannetti Vouet
Lanfranco
Patrons with at Least Eight Observations in the Estimated Sample
Aldobrandini Mattei
Altemps Mazarin
Barberini Medici
Borghese Orsini
Patrons with at Least Eight Observations in the Estimated Sample
Capocaccia Peretti montalto
Chigi Pointel
Church Pope
Colonna Roscioli
Filomarino Ruffo
Foreign nobles Sacchetti
Gonzaga Vatican St. Peter
King Valguarnera
Only Only
Patron FE Artist FE
(5) (6)
Still life (omitted)
Portrait 0.596 *** -0.263
(0.153) (0.371)
Genre 0.437 ** 0.031
(0.220) (0.390)
Landscape 0.864 *** 0.239
(0.180) (0.398)
Figurative (<5 fig) 1.268 *** 0.604 *
(0.172) (0.344)
Sacred < 5 (omitted)
History < 5 0.019 -0.223
(0.186) (0.292)
Myth < 5 0.072 0.014
(0.195) (0.201)
Figurative (5-10 fig) 1.636 *** 1.051 ***
(0.177) (0.347)
Sacred 5-10 (omitted)
History 5-10 0.009 -0.067
(0.226) (0.177)
Myth 5-10 -0.243 0.001
(0.282) (0.297)
Figurative (>10 fig) 1.858 *** 1.450 ***
(0.230) (0.387)
Sacred > 10 (omitted)
History >10 -0.175 -0.170
(0.243) (0.437)
Myth >10 0.380 0.181
(0.334) (0.361)
Battle >10 -0.941 *** -0.197
(0.252) (0.363)
Size 0.034 *** 0.038 ***
(0.007) (0.010)
Size^2 -0.00007 *** -0.00008 ***
(0.000) (0.000)
# commissions -0.099 *** -0.014
(0.019) (0.011)
Copy -0.996 *** -0.013
(0.225) (0.275)
Not on canvas -0.435 *** -0.217
(0.153) (0.196)
Fresco -0.472 ** -0.463 **
(0.204) (0.225)
Private chapel 0.797 *** 0 772
(0.240) (0.155)
Church and other rel 0.514 *** 0.891 ***
(0.159) (0.133)
Private palace 0.287 0.595 ***
(0.183) (0.147)
Private collection (omitted)
Minor destinations -0.105 -0.267
(0.220) (0.173)
Medium destinations 0.215 0.312 ***
(0.208) (0.117)
Exports 0.328 0.498 **
(0.271) (0.211)
Rome (omitted)
Age of artist 0.018 *** 0.017 ***
(0.003) (0.005)
Immigrant 0.117
(0.118)
Trend
Plague
Constant 1.136 *** 0.066
(0.258) (0.168)
Artist fixed effects No Yes
Patron fixed effects Yes No
Observations 716 716
[R.sup.2] 0.758 0.794
Artists with at Least Eight Observations in the Estimated Sample
Arpino
Baglione
Belloni
Both
Bril
Camassei
Caravaggio
Cerquozzi
Courtois
Cortona
V. de Boulogne
Domenichino
Dughet
Ferri
Gaulli
Gimignani
Grimaldi
Jannetti
Lanfranco
Patrons with at Least Eight Observations in the Estimated Sample
Aldobrandini
Altemps
Barberini
Borghese
Patrons with at Least Eight Observations in the Estimated Sample
Capocaccia
Chigi
Church
Colonna
Filomarino
Foreign nobles
Gonzaga
King
Patron &
Artist FE
(7)
Still life (omitted)
Portrait -0.349
(0.398)
Genre -0.030
(0.406)
Landscape 0.308
(0.441)
Figurative (<5 fig) 0.534
(0.370)
Sacred < 5 (omitted)
History < 5 0.003
(0.266)
Myth < 5 0.100
(0.205)
Figurative (5-10 fig) 0.912 **
(0.372)
Sacred 5-10 (omitted)
History 5-10 0.004
(0.205)
Myth 5-10 -0.354
(0.286)
Figurative (>10 fig) 1 183
(0.422)
Sacred > 10 (omitted)
History >10 -0.000
(0.538)
Myth >10 0.413
(0.358)
Battle >10 -0.282
(0.337)
Size 0.032 ***
(0.010)
Size^2 -0.00007 ***
(0.000)
# commissions -0.049 **
(0.024)
Copy -0.177
(0.221)
Not on canvas -0.280
(0.204)
Fresco -0.396 **
(0.193)
Private chapel 0.686 **
(0.302)
Church and other rel 0.645 ***
(0.153)
Private palace 0.557 ***
(0.174)
Private collection (omitted)
Minor destinations 0.109
(0.209)
Medium destinations 0.023
(0.165)
Exports 0.286
(0.218)
Rome (omitted)
Age of artist 0.018 ***
(0.005)
Immigrant
Trend
Plague
Constant - 1.742
(0.614)
Artist fixed effects Yes
Patron fixed effects Yes
Observations 716
[R.sup.2] 0.848
Artists with at Least Eight Observations in
Arpino 0.931
(0.568)
Baglione 0.839 ***
(0.252)
Belloni 1.072 ***
(0.198)
Both 0.327
(0.203)
Bril -0.017
(0.226)
Camassei 0.014
(0.306)
Caravaggio -0 721
(0.270)
Cerquozzi -0.409
(0.393)
Courtois 0.673 **
(0.293)
Cortona 0.264
(0.365)
V. de Boulogne 0.652 **
(0.256)
Domenichino 0.362
(0.222)
Dughet 0.654
(0.413)
Ferri 0.561 **
(0.218)
Gaulli -1.384 ***
(0.473)
Gimignani 0.157
(0.407)
Grimaldi -0.565 *
(0.312)
Jannetti 0.868 **
(0.402)
Lanfranco
Patrons with at Least Eight Observations in the Estimated Sample
Aldobrandini -0.203
(0.341)
Altemps -0.632 **
(0.307)
Barberini 0.868 ***
(0.252)
Borghese -1.388 ***
(0.335)
Patrons with at Least Eight Observations in the Estimated Sample
Capocaccia 0.239
(0.232)
Chigi -0.381
(0.399)
Church -0.178
(0.175)
Colonna -0.599 ***
(0.203)
Filomarino 0.097
(0.253)
Foreign nobles -2.098 ***
(0.373)
Gonzaga 0.815 ***
(0.202)
King -0.020
(0.320)
Notes: Robust standard errors in parentheses. Coefficients represent
deviations from artist-share weighted mean. Specification as in
column 7 of Table 3. FE, fixed effects.
*** p < .01, ** p < .05, * p < .1.
TABLE 4
The Components of Paintings' Prices
Total Observable Artist
Contribution Characteristics Effect
Observable characteristics 43.41 33.76 6.41
of which genre effect 2.52 -- --
Artist effect 24.20 6.41 23.34
Patron effect 17.16 3.23 -5.54
Error term 15.22 0.00 0.00
Total contribution 100.00 43.41 24.20
Patron Error
Effect Term
Observable characteristics 3.23 0.00
of which genre effect -- --
Artist effect -5.54 0.00
Patron effect 19.47 0.00
Error term 0.00 15.22
Total contribution 17.16 15.22
Note: Specification as in column 7 of Table 3.
TABLE 5
Robustness Checks
Only Multi- Only Multi- Match-
Genre Genre Specific
Artist Patrons Fixed Effects
Still life (omitted)
Portrait -0.100 0.192 -0.587
(0.454) (0.487) (0.498)
Genre 0.266 0.199 -0.090
(0.456) (0.328) (0.298)
Landscape 0.536 0.59 0.283
(0.492) (0.451) (0.523)
Figurative (<5 fig) 0.647 0.496 0.501
(0.455) (0.449) (0.456)
Sacred <5 (omitted)
History <5 -0.512 -0.022 0.094
(0.448) (0.393) (0.207)
Myth <5 0.225 0.357 0.222
(0.321) (0.223) (0.199)
Figurative (5-10 fig) 1.259 *** 1.142 ** 0.911 **
(0.478) (0.525) (0.452)
Sacred 5-10 (omitted)
History 5-10 0.093 0.359 0.32
(0.339) (0.357) (0.233)
Myth 5-10 -0.562 -0.113 -0.454
(0.460) (0.469) (0.318)
Figurative (>10 fig) 1.330 ** 1.236 ** 1.130 **
(0.538) (0.481) (0.475)
Sacred >10 (omitted)
History >10 -0.122 -1.098 ** 0.036
(0.767) (0.486) (0.651)
Myth >10 0.66 0.174 0.655 *
(0.443) (0.501) (0.342)
Battle >10 -0.616 - 1.014 -0.188
(0.484) (0.368) (0.330)
Constant 0.485 - 1 -2.190 ***
(0.819) (0.658) (0.782)
Observations 343 160 716
[R.sup.2] 0.849 0.901 0.894
Notes: Robust standard error in parentheses. Specification as
in column 7 of Table 3 including match-specific fixed effects.
*** p <.01, **p < .05, * p <.1.
