Revisiting Harrison and Cynthia White’s Academic vs. Dealer-Critic System

: The ﬁeld of art market studies is based on a famous opposition, coined by Harrison and Cynthia White in 1965, regarding the “academic” system, as opposed to the “dealer-critic” one. Published in 1965, their book, Canvases and Careers, Institutional Change in the French Painting World , was qualiﬁed by Patricia Mainardi and Pierre Vaisse, but their criticism dated back to the 1990s. In the meantime, the development of digital methods makes possible a broader reassessment of Harrison and Cynthia White’s theory. Based on a corpus of Parisian auction sales, from 1831 through 1925, this paper uses econometrics to call into question the antagonism between the academic and the dealer-critic system, and comes to another conclusion: the academic system was crucial to determine the value of artworks and its efﬁciency did not collapse in the 1870s, nor in the 1880s, but rather after the Great War.


Introduction
In the field of art market studies, Harrison and Cynthia White's book Canvases and Careers, Institutional Change in the French Painting World, appears as a foundation stone for any paper. Published in 1965, it analyzes the artists'careers in the 19th century and describes the transition from one institutional system to another, i.e., from the academic system to the dealer-critic one. The former, heir to the Ancien Régime, is based on the Academy of Fine Arts, the School of Fine Arts and the Salon. To make a career, the artist had to be received at the School of Fine Arts, enter the spiral of competitions-the Rome Prize was the ultimate stage-and make himself known to the public by participating in the Salon. This exhibition was a token of recognition: a jury decided on admissions and then rewarded the "best" artworks with first-, second-and third-class medals. 1 The Salon captured the attention, caused ink to flow in the press and also constituted a commercial place where the State and collectors could buy the works on display. It was even the main supply source of artworks for the French public authorities.
The academic system showed signs of running out of steam from the late 1840s, with the growing number of artists who were accepted at the Salon: this exhibition became saturated and, from the point of view of an exhibitor, too large to be noticed. Moreover, the jury had become unpopular, due to its severity and to the opening of a Salon des refusés in 1863. From then on, the "dealer-critic system triumphed, in part because it could and did command a bigger market than the academic-governmental structure" (White and White 1965, p. 98). Harrison and Cynthia White situated the transition from 1 The Salon was biennial from 1798, annual from 1833 to 1851, again biennial from 1853 to 1863, then annual from 1864 (Monnier 1995, p. 124).
the academic system to the dealer-critic one in the 1870s, with the Impressionists 2 and the dealer Paul Durand-Ruel, who appeared as a patron, because of the regular support he offered to artists. Unlike the academic system, which individualized careers by focusing on the canvases presented at each Salon, its rival had the advantage of building longer-term careers and, through group exhibitions, it encouraged pools of artists.
The transition from the academic system to the dealer-critic one continues to be a consensus among the researchers who focus on this institutional change. 3 Pierre Vaisse found that Harrison and Cynthia White "have raised the problem of the Salon in relevant terms" (Vaisse 1995, p. 96). As for Patricia Mainardi, she postponed their chronology by only a few years: after the French state abandoned the Salon to the artists' control in 1880, "the period from 1880 to 1885 saw the termination and collapse of the official French Salon system" (Mainardi 1993, p. 129). 4 Finally, in 2002, David W. Galenson and Robert Jensen wrote a large paper in which they pointed out their disagreements with the Whites, whose analysis "credits the wrong people for changing the art world in the 19th century" (Galenson and Jensen 2002, p. 46) but they did not revise the chronology, stating that "the official Salon had been the key institution in the French art world until 1874", i.e., until "the impressionist exhibitions broke the Salon monopoly over the ability to present artists' works to the public in what would be generally considered as a legitimate setting" (Galenson and Jensen 2002, pp. 35, 42).
To establish evidence of a collapse of the Salon in the 1870s-or 1880s-Harrison and Cynthia White analyzed a random sample of a dictionary of artists (White and White 1965, p. 111), and Patricia Mainardi relied on speeches and debates about the Salon. While these sources are valuable documents, they are not sufficient, however, to show that the Salon ceased to play a role in artists' careers: the first is not exhaustive and the second does not concern artists' careers. The challenge of this article is to measure the efficiency of the Salon by measuring the artists' economic performance at Parisian auctions. The conclusions then differ widely from those of Harrison and Cynthia White: only after the Great War did the official Salon cease to have an impact and did the academic system collapse.

Materials and Methods
To test the weight of the Salon on artistic careers, it would be ideal to have an exhaustive biographical information on all artists and to link it to the economic performance of their works. However, the artist dictionaries used by the Whites do not contain any exhaustive information on the prices of artworks. One solution to this methodological problem is to use a source that provides access to the price of works over a long period of time. Parisian auctions are the only way in. The results presented in Section 3 are thus based on the analysis of a twofold dataset, regarding Parisian auctions of paintings, drawings and sculptures.
2.1. The Exhaustive Surveys of 1875, 1900 and 1925 Auction catalogues are a very valuable source that allows a macroscopic view, whereas an artist's or dealer's account book only considers part of the market, by definition. The Getty Research Institute has put online a database of annotated catalogues, mainly for the 18th century up to the 1820-1840s in Europe, and for the 1900-1945 period in Germany. This exceptional source has already made it possible to identify the major changes in London sales between 1780 and 1835 (Fox and Lincoln 2016). It should be noted, however, that the period 1840-1900 is absent from the Getty's dataset, which hinders any 2 "The dealers and critics, one marginal figures to the Academic system, became, with the impressionists, the core of the new system." (White and White 1965, p. 151). 3 The bibliography of the art market is very abundant. Nevertheless, the majority of publications do not take this institutional change as the focus of their study-for example, monographic studies of dealers, such as Goupil, Durand-Ruel, Vollard or Kahnweiler. Robert Jensen analyzed the socio-economic profile of dealers (Jensen 1994) and Béatrice Joyeux-Prunel, that of the avant-gardes (Joyeux-Prunel 2009). 4 This two-fold source is the only one that provides a comprehensive dataset. The private sales between dealers, collectors and artists thus constitute a blind spot, as there is no available and global archive.
cross-sectional analysis of the market for the 19th century. To have an overall view of Parisian auctions, it is, therefore, necessary to group together all the auction catalogues regarding paintings, drawings and sculptures and to transcribe them by hand into a database: this task being impossible for a single person over the entire period, I carried out surveys by considering all the auction catalogues for certain years: 1831, 1850, 1875, 1900, and 1925. 5 In this paper, only the 1875, 1900, and 1925 samples are used to support the analysis.
After collecting the auction catalogues, 6 I fully transcribed them into a spreadsheet, separating all the fields regarding the auction sale in different columns 7 and extending this transcription to all the objects in the catalogue. 8 In a second step, I matched the auction catalogues with their minutes ("procès-verbaux"), when available at the Paris archives. The minutes are a unique and typically French source, attesting to the transfer of ownership of each lot. 9 By definition, a "procès-verbal" is an authentic and enforceable deed, used as a means of proof in case of a dispute or a claim. Officially drawn up by the auctioneer who carries out the sale-de facto, by his clerk-it mentions the following information: the name, address and occupation of the person at whose request the sale is made, 10 the summary description of each item put up for sale with the hammer price in full, and finally, the name, or even the address, of the winning bidder. 11 As the minute is produced by the auctioneer himself, there cannot be any errors on the hammer price or the identity of the purchaser, unlike annotated catalogues. It is therefore the most exhaustive and reliable reference source available for researchers.
The combination of the auction catalogues and the corresponding minutes therefore provides information on the artworks and their hammer price for all the paintings, drawings and sculptures put up for sale in Paris in 1875, 1900 and 1925. At the end of this long work, this corpus includes 11,586 paintings, drawings and sculptures for sale in 1875, 8,137 in 1900 and 24,724 in 1925. The bibliographic references are fully available online (Saint-Raymond 2018c), as well as the database that served as a basis for the econometric analysis (Saint-Raymond 2018f).
Like most economic papers on the art market, seeking to explain the price of works, 12 I considered the following explanatory variables. First, I isolated the characteristics of the artwork sold at auction, with medium dummies 13 (painting/sculpture/graphic work), size variables and the presence of a 5 I started in 1831 when the total revenue of the Parisian auctioneers become available; I then moved to 1850; taking a sample in 25 year blocks thereafter. 6 The auction catalogues were collected at the INHA library, at the Département des Estampes of the Bibliothèque nationale de France, on gallica and at the library of the Ecole nationale des beaux-arts. The choice of catalogues is based on the "long" title, and not on the short title of the first cover: this is why I had to carefully examine all the catalogues concerning the years of interest, without stopping at the library search engines which, generally, only take the short title into account. 7 I separated the following fields into different columns: date and venue of auction sale, names and addresses of the auctioneer and expert, short and long title of the catalogue, name of the seller, amount of buyer fee, total number of lots, number of pages in the catalogue, possible presence of illustrations, presence of a preface and name of any prefacier, name and address of the publisher. 8 Regarding the artwork, I separated the following fields into different columns: section of the catalogue in which the artwork was sold, catalogue number, name of the artist, title, date, presence of a signature, dimensions, medium, comprehensive description of the work, presence of a reproduction, exhibitions in which it appeared, possible exhibition of the work at the Salon, previous auction sales and collectors-this information forms the "pedigree" of the work-, mention of an engraving of the work and presence of a bibliography documenting the object. The catalogue annotations, if any, were added in separate columns, for the hammer price, the purchaser and, more rarely, the price setting and estimation. 9 In the other countries, the auctioneers were not considered as ministerial officers, i.e., State representatives, subject to very strict legislation. In England, for instance, there are no regulations on auctions (Maurice 1971, p. 41-56). 10 The "real" seller can use an intermediary: in this case, the minute of the sale would only mention the name, address and occupation of the latter. Regarding the auction sales with different sellers-"composite sales" or "ventes composées"-, please confirm/ LSR : yes, necessary. the list of all the sellers appeared in the minutes from the 1920 onwards: beforehand, only the name of the expert was mentioned. 11 The information on the winning bidder's identity was not mandatory in the minutes, for instance it was absent from the minutes of the Bordeaux auction sales (Rouillon 1928, p. 109). 12 See, for instance, (Mei and Moses 2002;Ashenfelter and Graddy 2003;Renneboog and Spaenjers 2012;David and Oosterlinck 2015). 13 A dummy variable equals to 1 if the proposition is true (e.g., the artwork is a painting), 0 otherwise. signature. 14 Attribution was also taken into account, based on the description of the artwork in the auction catalogue. I thus distinguished three dummy variables: artworks "made by" the artist, artworks "attributed" to the artist and, finally, artworks in the "style of" the artist (grouping together the mentions "school of" the artist/"in the style of"/"after" the artist). The "pedigree" variable equals 1 if the auction catalogue mentioned a previous collector or a previous auction sale for the artwork, 0 otherwise. Furthermore, the "Salon" variable equals 1 if the auction catalogue stated that the artwork was shown at the Salon. It differs from the "Exhibition" variable, which follows the same principle, but for all exhibitions other than the official Salon. 15 I also paid attention to the information given by the catalogue: the "Bibliography" variable indicated a bibliographic reference for the artwork that was sold at auction, and the "Engraved ?" dummy variable indicates whether the work was previously engraved. However, contrary to all the other papers in art market studies, my econometric results also include two variables on the auction catalogue itself: the length of the description for the artwork, i.e., the number of words, and a "Reproduced" dummy that equals to 1 if the work was reproduced in the catalogue, 0 otherwise. Finally, the "Sold by the artist?" variable indicates whether or not the artwork was sold by the artist himself.

Auctions Organized to Their Artists for Their Own Works
The constitution of this second corpus followed a more complex process than the previous one: it was no longer a question of grouping together catalogues of paintings, drawings and sculptures, but of identifying the sales organized by artists for their own works.
The main source of this corpus is Frits Lugt's repository of auction catalogues (Lugt 1938), which ends up in 1925. First, all the Parisian auction catalogues with the mention "by him", "by her" or "by them", were selected: this means that the works put up for sale in Paris were made and sold "by" the artist. I extended the selection to the entries in which the seller's occupation was "artist" or "painter" or "draftsman" or "sculptor". This corpus was completed by the "quitus" 16 of the Parisian auctioneers, kept in the Paris archives including sales for which the claimant was a "known" artist or already mentioned in the previous stage.
To ensure that the sales were indeed organized by the artists who wanted to sell their own works at auction, it was necessary to ensure that these artists was indeed alive at the time of the sale, i.e., that the auction sale wasn't actually the liquidation of their studio, after death. I therefore identified the artists thanks to the Bénézit dictionary (Bénézit [1911(Bénézit [ ] 1999. Once these biographical checks were carried out, I collected all the catalogues and eliminated the auction sales that did not concern the artist's production as such, but rather any movable effects (furniture, personal objects, etc.). Then, I needed to check the identity of the seller, 17 so used two sources: the minutes of the sale and, if the minutes of the sale are unavailable, the quitus of the auctioneers, accessible from 1864 onward.
In the end, this final corpus is divided into three groups, according to the degree of certainty of the seller: (1) sales where the applicant was indeed an artist or group of artists; (2) sales where the applicant was a dealer or expert, as it is impossible to verify whether the dealer/expert auctioned off his own stock or whether he acted as an intermediary or a mere applicant for the artist, and (3) uncertain sales, for which neither the minutes nor the quitus are available. I enriched this dataset with some biographical data, through the Bénézit dictionary. In particular, I added the artists' age at the moment of the sale, and a dummy variable that equals to 1 if the artist is a sculptor. If the artist previously organized an auction sale for his/her works, the variable "already come" equals to 1. I made a second data enrichment, regarding the artists' activity at the Salon. To do so, I used the "Base Salons et expositions de groupes 1673-1914", developed by the Musée d'Orsay and INHA. This resource is exhaustive until 1879: for the period after 1900, no online database provides any comprehensive information. Before the creation of the Société des Artistes français, in 1882, I thus created the variable "Experience", accounting for the total number of years since the artist first exhibited at the Salon. The dummy variables "hors concours" and "medal" respectively equal to 1 if the artist was awarded as a "hors concours" or received a medal at the Salon. After 1882, I therefore chose not to enrich my auction corpus with external information, but to analyze the auction catalogues themselves: the variable entitled "mention of the Salon" thus indicates whether the catalogue highlights the Salon in the artist's career or in the description of the works. Through the whole period, the variable "nb of lots" gives the total number of artworks that were sold at auction, and "illustrated cat." equals to 1 if there were any reproduction of artworks in the auction catalogue.
This corpus and the data that I used for the econometric analysis are available online (Saint-Raymond 2018e).

Econometric Analysis of These Two Datasets
These two corpora were analyzed by using the hedonic regression method. The purpose of this technique is to describe how an "explained variable" varies according to different "explanatory variables". In this paper, the two explained variables are the hammer price 18 (Table A1) and the average sales revenue (Table A2). The hedonic regression method remains descriptive: it highlights correlations but does not reveal causalities. 19 Its advantage is to make comparisons "all other things being equal", i.e., pretending that all the observed characteristics are unchanged, except for the explanatory variable in question. For instance, the impact of the explanatory variable "reproduction in the catalogue" is calculated as if two works were identical in all respects, with the difference that one will be reproduced in the catalogue and the other not. The correlation of a variable to the auction price can then be isolated by "freezing" all other explanatory factors. If the explanatory variable has an effect on the explained variable, then it is said to be "significant". The R 2 is called "coefficient of determination": it gives the proportion of the variance in the explained variable that is predictable from the explanatory variable(s). For instance, in the regression model 1 (Table A2), R 2 equals to 0.2951, i.e., the explanatory variables of this model account for 29.51% of the variation of the average sales revenue, between 1870 and 1881.
In the results tables displayed in the appendices, each explanatory variable corresponds to a line, and a figure, or coefficient, corresponds to the "direction of correlation". If the number is positive, then the correlation between the explained variable and the explanatory variable varies in the same direction. For example, in regression model 1 (Table A2), the variable "Nb of lots" is negatively correlated with the average sales revenue because the coefficient is negative, equaling to −1.04. In the case where the variable explained is quantifiable-like the auction price of the average sales revenue-then the coefficient is also a number. In the last example, one extra artwork, sold at auction by the artist, makes the average sales revenue decrease by 1.04 francs, all other things being equal. By comparing the coefficients, we can thus compare the effects of the different variables on the auction price.
Nevertheless, the reading should not be stopped at simple coefficients: it is necessary to test the accuracy of the variables. To do this, the figures in brackets indicate the standard deviation of each coefficient, and the number of stars, its significance. Without going into detail, a coefficient is significant if its standard-deviation around it is "not to high". Three stars mean that the coefficient is "true"/significant at a 99% level; two stars-respectively one star-mean that the accuracy of the coefficient is 95 to 99%-resp. 90 to 95%. When the coefficient has no star, its standard deviation is too high, so there is no real correlation between the explanatory and the explained variables.
For more rigor in the regressions, it is possible to add a "fixed effect", i.e., to perform this regression technique by separating various entities, or groups. 20 For example, with an "artist fixed effect", the regressions are performed by considering each artist separately and not altogether. The results are thus interpreted for a given artists, all the other things being equal.
The regressions were performed with the STATA software. The programming commands have been mentioned in the files available online, to ensure the reproducibility of the results.

Results
The econometric analysis of these two datasets, as shown in Tables A1 and A2 (Appendix A), sheds light on a different chronology than Harrison and Cynthia White'.

The Salon And the Market: Complementary and Non-Rival Systems (1850s-1870s)
With a demographic growth of art dealers in Paris from the 1850s (De Maupeou and Saint-Raymond 2013) and the increase of auction sales, following the opening of the hôtel Drouot in 1852 (Saint-Raymond 2018a), the Parisian art market experienced an unprecedented boom. For Harrison and Cynthia White, the dealer-critic system appeared as a rival of the Academic one, the former replacing the latter between the 1850s and the 1870s (White and White 1965, pp. 2, 151). However, the expansion of the Parisian art market does not necessarily mean that the Salon stopped acting as a taste-maker and gate-keeper.

Evidence from the Global Econometric Regression
Many monographic studies in art history highlight the role of the Salon as a springboard, allowing an artist to gain a certain notoriety and, in a second step, to make himself known by dealers. Félix Ziem's career is particularly representative of this phenomenon (Saint-Raymond 2016a) as well as the Italian orientalist painter Alberto Pasini: the latter began to exhibit at the Salon from 1857, he won third-class medals in 1859, second-class in 1863 and first-class in 1864: from these years on, he entered into a commercial relationship with the dealer Adolphe Goupil, who monopolized his production, took care of his sales and supported his prices at auction (Saint-Raymond 2018b). Another significant example is the success of Alphonse de Neuville's painting, Combat sur une voie ferrée (Figure 1): exhibited in the 1874 Salon, it was sold at auction one year later and bought for 11,500 francs by the lawyer Georges de Chambine: this hammer price ranked Alphonse de Neuville among the most valued living artists at Parisian auctions in 1875 (Saint-Raymond 2018c). 21 20 The R 2 within gives the part of the intra-individual variability of the left variable explained by those of the right variables.
The R 2 between estimates the contribution of the fixed effects to the model. The R 2 overall reflects the overall quality of the regression. 21 Lugt 35571, lot 56. The bibliographic reference of the auction catalogue is summarized by its number in the repository by Frits Lugt (1938), and it is available online (Saint-Raymond 2018d). Beyond these specific examples, which are difficult to generalize, it is possible to globally measure the correlation between the presence at the Salon and the economic performance of an artist or an artwork. A good test is to consider all Parisian auctions of paintings, drawings and sculptures and to test the correlation between their auction price and the fact that the artworks were previously exhibited at the Salon-this information is always mentioned in the catalogues, which became very detailed in the 1870s. The year 1875 is an interesting date: the first impressionist exhibition took place a year earlier and in March 1875, the impressionists held their first auction sale in Paris. For Harrison and Cynthia White, this would therefore be the end of the academic system. Table A1 clearly contradicts their thesis. Indeed, the mention of the Salon significantly increased the hammer price of the work (excluding repurchases): all other things being equal and for a same artist, the hammer price increased by 2528 francs when the artwork was previously exhibited at the Salon, which is very important compared to the other variables. 22

Evidence from the Auction Sales Organized by the Artists Themselves
On the other hand, it is possible to measure the influence of the academic system on artistic careers, based on an experience that artists themselves had, supposedly, to free themselves from the yoke of the Salon and its jury: the organization of an auction sale of their own works. This practice flourished in the period 1852-1881. 23 Until the 1840s, artists reluctantly went to the auction house, citing departure or health reasons to publicly justify the auction sale. The upheavals of the Salon, under the Second Republic, had then led to a first influx of artists, mainly landscape painters, who willingly put their works up for auction (Saint-Raymond 2018e). After 1852, with the opening of the hôtel Drouot and the subsequent expansion of the public, more and more artists organized an auction sale, during two waves, the first one in the late 1850s, of a magnitude equal to that of 1849-1851, the second one from 1867 onwards, culminating in the late 1870s. The auctioneer Jules Auguste Boussaton played a very important role in stimulating these artistic sales. 24 ) He was at the center of the network of professionals (Figure 2) who organized these events, along with some Parisian dealers who also encouraged "their" artists to sell their works at auction-especially those who held an expert position, such as Pierre-Firmin Martin and Paul Durand-Ruel. 25 The artists who went to the hôtel Drouot were almost all men 26 and, mainly, landscape painters (Saint-Raymond 2018e). They asserted themselves in a more positive way than in the 1840s, as shown by the semantic change in the catalogue prefaces: they no longer "got rid of" their works, but "offered" them to the public or, in a more agonistic register, "delivered" them to the fire of the auction. Indeed, with its larger clientele, the new auction house-the hôtel Drouot-was then perceived as the place where artists could really "enter into communion" 27 with the audience. As the critic Philippe Burty stated, "the hôtel Drouot alone offers the advantage of attracting crowds on a daily basis and therefore provoking criticism to the fullest". 28 Behind this romantic and mythical vision of the artist, being in direct contact with the amateur, there was a virulent criticism of the Salon which, by the "excessive agglomeration" 29 of the works on display, plunged the painters and sculptors into anonymity (Vaisse 1983, p. 78). The hôtel Drouot thus had a double advantage over the Salon: it was much more democratic and open, without the risk of a refusal or prohibition and without the barrier of the jury's "firing squads", 30 and it gave the artist better visibility, by offering the possibility of a real solo exhibition, 31 with a substantial body of works and not "the two or three artworks tolerated annually by the jury". 32 Was the hôtel Drouot the salvation plank of young artists, in order to launch artistic careers and to escape the Salon's sanction? The econometric analysis for the period 1870-1881 (Table A2, model 1) shows that the average revenue of the auction sale was all the more important as the artist had succeeded in reaching the rank of "out of competition" ("hors concours") at the Salon, meaning that he was exempted from going before the jury. Indeed, between 1870 and 1881, the average income from sales organized by artists for their own works was significantly higher if the artist was "out of competition": it increased by 201 francs, all other things being equal. On the contrary, obtaining third-, second-or first-class medals had no significant impact on the auction price (Table A2, model 2).
Two other variables had a significant impact on the average sales revenue. On the one hand, the sculptors succeeded much better than their comrades, all other things being equal. On the other hand, and not surprisingly, the average auction price decreased with the quantity of works auctioned. It was therefore more profitable to sell little and control the scarcity of the artworks, than to flood the market.
These statistics thus contradict Harrison and Cynthia White' theses: the dealer-critic system did not replace the academic one in the 1870s. On the contrary, the Salon continued to have a very heavy impact on the artists' careers, even for those who wanted to escape it through auctions. The artists themselves were not mistaken. While criticizing the academic system, they continuously highlighted their awards at the Salon, in the title of the auction catalogue, 33 the preface 34 or pedigree of the works, even if it meant embellishing reality. 35 The landscape painter Louis Adolphe Hervier and the history painter Joseph Navlet embodied some unfortunate examples: despite their efforts to make a career at auction 36 and, for the first, despite the support of critics such as Théophile Gautier and Philippe Burty, 37 they cruelly missed the honors of the Salon. Both obtained catastrophic results at each of 26 Only Louise Adèle Dumoulin, Louise Daru and Berthe Morisot sold their own works at auction between 1852 and 1881. 27 Lugt 34703,35492,35496 and 35521. 35 In 1876, Ernest d'Hervilly stated, in the preface to the sale of Jules Héreau, that the artist had obtained "for ten years the medals of the official Jury" (Lugt 36364, p. 1) but the Salon catalogue, this year, did not specify any official reward. 36 Hervier organized seven sales, from 1856 to 1878, and Navlet, ten, between 1869 and 1885. 37 Théophile Gautier wrote an article in the Moniteur universel of 11 February 1858, which was included in the preface to the sale of 5 April 1875 (Lugt 35516) and Philippe Burty wrote the preface to the following sale (Lugt 36195). their sales, most of which were less than 50 francs per work, and barely enough to cover the costs (Saint-Raymond 2018e).

The Collapse of the Salon System? (1880s-1910s)
The threshold of the 1880s marked a turning point in the history of the Salon: with the decree of 27 December 1880, the State entrusted all French artists with the free and complete, material and artistic management of the Salon. A "Société des Artistes Français" was set up on 20 March 1882, in order to organize the annual Salon, i.e., the public and official exhibition of living art (Maingon 2009, p. 134). However, this same year, this event lost its cohesion, as the State decided to set up a-quickly abandoned-triennial Exhibition in parallel. In 1884, a Salon des "indépendants" appeared on the fringes of the annual Salon. Lastly, in 1889, a dispute led to a split between the Salon de la Société des Artistes Français and a newly created Société nationale des Beaux-Arts.
If Patricia Mainardi analyzes the institutional crisis of the unique Salon as a collapse of the latter, which paralleled political and economic tensions within the Third Republic (Mainardi 1993), Pierre Vaisse is more measured: not only did the official purchases and orders of the State continue but the "Official Salon" remained the Salon des Artistes Français, in the eyes of contemporary artists (Vaisse 1995, p. 94).
However, one question remains open: did the Salon des Artistes Français really lose all influence on the artists' careers? Did it only remain a façade, with no real effect on market? By studying Parisian auctions, it seems that the Salon continued to play a major role in determining prices.

Evidence from the Global Econometric Regression
Following the same methodology as in 1875, the regression coefficient for the "Salon" variable, in 1900, was still positive and very significant: all other things being equal, and for a same artist, the hammer price of a work in 1900-excluding repurchases-increased by 16,210 francs if it had previously been exhibited at the Salon (Table A1). This result alone therefore calls the Whites' theses into question because it shows that in 1900, the Salon played a significantly strong role in determining prices.
Table A1 also proves that, while the academic system remained a major prescriber in auction sphere, the dealer-critic system was also gaining in importance. Indeed, when the work put up for sale had been exhibited in a dealer's gallery, its hammer price increased significantly but 5 times less than if the work had been exhibited at the Salon. Similarly, the price of a work was positively correlated with the way it was highlighted in the auction catalogue-by mentioning a bibliography on it, the number of words in the description and, above all, a reproduction by photoengraving. This enhancement through increasingly luxurious catalogues, was encouraged by the dealer Georges Petit (Figure 3), who played a crucial role in the Parisian art market.
At the age 25, Georges Petit succeeded his father and, inspired by the Grosvenor Gallery in London, he opened a very luxurious new exhibition venue, in 1882, at 8 rue de Sèze. He proposed this gallery to the Parisian auctioneers, for some prestigious auctions sales-of which he would also be an expert. The first one took place in April 1883. Georges Petit continued this model of vertical integration, based on luxury, by founding his own printing house in 1894. The auction catalogues he published were truly beautiful books, prefaced by art critics and illustrated by full-page reproductions, using the heliogravure printing process. In about ten years, Georges Petit revolutionized the Parisian dealer-critic system, since he invented a new "model" for the enhancement of works (Saint-Raymond 2018g). As shown by the econometric analysis (Table A1), this model had a very strong influence on hammer prices.

Evidence from the Auction Sales Organized by the Artists Themselves
If, in 1900, the Salon maintained a positive effect on auction prices, was it still as important in determining artistic careers? It is possible to measure the emotional and symbolic link maintained by the artists with the Salon, by observing the prefaces of the catalogues, when the artists sold their own works at auction. In the 1880s, they maintained a strong presence at the hôtel Drouot. Then, from the 1890s onwards, there were less than ten auctions per year, until a very marked slowdown in the 1910s (Figure 4). 38 Despite this decline in volume, art critics continued to preface many auction catalogues. The textual analysis of these prefaces also goes against a so-called collapse of the Salon: between 1880 and 1912, half of the prefaces mentioned the Salon des Artistes Français, in order to highlight the careers of artists who sold their works at auction ( Figure 5).
Moreover, this graph shows no trace of any symbolic collapse of the Salon over time: whatever the period, half of the auction prefaces mentioned the Salon. In 1910 and 1912, the two available prefaces insisted on the role of the official Salon in the careers of Albert Ferdinand Duprat and Hyacinthe Royet. For artists, the Salon therefore remained a means of distinguishing themselves and a guarantee of their professional "quality". The example of the painter Euphémie Muraton is meaningful: when she put her still-lifes up for auction on 26 March 1888, she had an illustrated catalogue printed, prefaced by the journalist Octave Robin. The latter began his text by explaining how he discovered her works at the Salon: whose artistic reputation spread so brilliantly. My companion considered this exquisite canvas for a long time, then he said to me: "Here is some painting! [. . . ] See how all this sings in a discreet and fair harmony. This is one of the best things at the Salon." I found this painting among those of Mrs. Muraton who were selected for this sale, and it seemed even more charming to me. 39 The preface begins with the work that also opens the auction catalogue; exhibited at the 1887 Salon Old Memories benefited from full-page reproduction ( Figure 6). This symbolic enhancement, by the mention of the Salon and by the engraving, had an economic consequence: Old Memories obtained the highest auction price of the sale, bought 1320 francs, excluding buyer fee, by a certain Mr. Piegu. 40 The example of this work cannot be generalized, due to the lack of a sufficient ratio of works exhibited at the Salon and sold by artists at auction. If it is difficult to carry out regressions at the level of works, it is nevertheless possible to measure the impact of the Salon on sales, by analyzing the correlation between the mention of the Salon in the preface, and the average sales revenue, in constant francs.
All other things being equal, the mention of the Salon in the preface significantly increased the average sales revenue for artists who put their works on sale between 1882 and 1912, by 105.5 constant 1890 francs per lot (Table A2, model 3). The revenue was also rising if the catalogue was illustrated. This result is striking and goes in the same direction as that of Table A1: the Salon therefore continued to play a positive role in the price of works and the careers of artists, economically and symbolically. In other words, the artists who entered the arena of auctions met a success all the more important as they had been recognized at the Salon, between the 1880s and the early 1910s.   1 8 9 0 1 8 9 1 1 8 9 2 1 8 9 3 1 8 9 4 1 8 9 5 1 8 9 6 1 8 9 7 1 8 9 8 1 8 9 9 1 9 0 0 1 9 0 1 1 9 0 2 1 9 0 3 1 9 0 4 1 9 0 5 1 9 0 6 1 9 0 7 1 9 0 8 1 9 0 9 1 9 1 0 1 9 1 1 1 9 1 2 Auction catalogue whose preface mentions the Salon Auction catalogue whose preface does not mention the Salon, but selling artworks previously exhibited at the Salon Auction catalogue whose preface does not mention the Salon Source and visualization : L. Saint-Raymond, 2019. N = 73 prefaced auction catalogues, regarding auction sales that were organized by the artists themselves, for their own works, from 1882 through 1925.

The End of the Salon, after the Great War
After the Great War, artists nearly stopped organizing auction sales for their own production ( Figure 4): it thus becomes impossible to carry out any econometric analysis for this corpus. However, the collapse of the official Salon can be tested through the exhaustive survey carried out on all the Parisian auction sales of paintings, drawings and sculptures in 1925 (Table A1). Unlike in 1900, the mention of an exhibition at the Salon des Artistes Français-the official Salon-ceased to play a significant and positive role in the auction price (Table A1). On the other hand, the other exhibitions were positively correlated with the price-among them, the Salon d'Automne and the Salon des Tuileries.
After the war, the Salon des Artistes Français thus crumbled. Indeed, the mention of the Salon had almost completely disappeared from the auction catalogues: in 1925, only eight works sold at auction were previously exhibited in a Salon des Artistes Français, after 1918, among which six paintings by the Egyptian painter Georges-Hanna Sabbagh. This mention did not bring success to the artists 41 and Sabbagh had to buy three works back, at a very low price, below 250 francs. 42 The journalist and art critic André Fage confirmed these results: in 1930, the opening of the official Salon "went almost unnoticed in the worldly splendour" and the paintings of these artists "were generally sold very cheaply at the hôtel Drouot", while "the works of most of the painters of the Salon d'Automne, the Salon des Tuileries and the Independents, were sold at a golden price on auction sale." (Fage 1930, pp. 232-34). More than the Salon des Tuileries and the Salon des Indépendants, which were rarely mentioned in the 1925 auction catalogues, the Salon d'Automne seemed to have a stronger power of prescription for young painters. La Femme couchée by Maurice Asselin, and Les Bucoliques by Jules Léon Flandrin, both exhibited at the 1921 Salon d'Automne, reached high prices in 1925, respectively 6050 current francs, excluding buyer fees, to Baron Fukushima and 5900 francs to the Druet gallery. 43 If the Salon des Artistes Français-and the academic system-had lost its efficiency, it meant that the dealer-critic system had become all-powerful to build artistic careers. This supremacy can be read on two levels. First, there was a parallel between the ranking of the most valued artists at auction, on the one hand, and the ranking of the most exhibited artists in the dealers' galleries, on the other hand (Saint-Raymond 2019). As the number of Parisian galleries exploded after the war, rising from 146 dealers in 1918to 195 in 1929(De Maupeou and Saint-Raymond 2013, the exhibitions also experienced an exponential growth. Second, the omnipotence of the dealer-critic system was made visible with the disappearance of individual auction sales organized by the artists themselves for their own works. Indeed, the multiplication of exhibitions (in the dealers' galleries) offered a springboard for young artists, making the auction scene lose its role as a substitute.

Discussion
This research allows qualifying Harrison and Cynthia White's, then Patricia Mainardi's, chronology: instead of showing a transition from the academic system to the dealer-critic one in the years 1870-1880, our econometric results prove that the official Salon continued to play its role as a prescriber until the early 1910s. Artists entering the auction arena were all the more successful as they had been recognized and rewarded at the Salon. In the aftermath of the Great War, the disappearance of the academic system as a legitimizing body, gave the dealer-critic system its full place: the consecration of artists, thus determined "from within" by dealers, critics, collectors and museums, was interpreted as the triumph of finance, or even the shameless agiotage of speculators-even today, this absence of "counter-powers" to the dealer-critic system gives rise to virulent denunciations, of a contemporary art dictated by the logic of speculation and money.
These results are largely dependent on econometric analysis of the 1875, 1900 and 1925 surveys. A research team would be needed to extend the research to other years and thus further refine the chronology. As this paper is based on a very specific part of the art market-auctions-it thus constitutes a first step to extend this methodology to private sales between dealers, collectors and 41 These eight paintings did not exceed 500 francs. Lugt 88024 lots 49-53, Lugt 89502 lot 134 et Lugt 89394 lot 141. 42 Lugt 88024, lots 52, 54 and 40, repurchased respectively 150, 205 and 250 francs excluding fees. 43 Lugt 88156,lot 8. Lugt 88121,lot 84. artists. Unfortunately, the scarcity of accounting books makes this research very difficult to undertake at a global level.
Funding: This research received no external funding.

Conflicts of Interest:
The author declares no conflict of interest.
Appendix A Table A1. Results of the hedonic regression, with a fixed "artist" effect, conducted on Parisian auction sales of paintings, drawings and sculptures, in 1875, 1900 and 1925. The explained variable is the hammer price (in current francs), excluding bought in prices (i.e., repurchases) and excluding fees.