Research Center

Despite  the  existence  of  a  large  literature  dedicated  to  the  art  market,  the  management  of  art galleries  remains  a  black  box  as  both  buyers  and  sellers  put  a  high  value  on  discretion.  This paper overcomes the traditional lack of quantitative data by analysing the complete books of one of  the  most  successful  French  galleries  at  the  end  of  the  19thcentury:  Goupil,  Boussod  & Valadon.  This  original  database  covers the  sale  of more  than 40,000artworks  that  occurred between 1860  and 1914. Rapidity to  sell  artworks  was  a  key  element in  Goupil’s strategy.  As  a whole  Goupil  sold  71%  of  artworks  bought.  Out  of  the  sold  artworks,  almost  80%  were  sold within  a  year  (with  22%  before  the  acquisition  reflecting  orders).Goupil  extracted  a  higher return  from aristocratsand  competitors  whereas  branches  and  large  buyers  were  getting  a discount. Artworks  from  artists  in  residence  were  on  average providing  lower  returnsbut  were less likely to be sold at a loss.

In this article we make a case for a systematic application of complex network science to study art market history and more general collection dynamics. We reveal social, temporal, spatial, and conceptual network dimensions, i.e. network node and link types, previously implicit in the Getty Provenance Index® (GPI). 1 As a pioneering art history database active since the 1980s, the GPI provides online access to source material relevant for research in the history of collecting and art markets. Based on a subset of the GPI, we characterize an aggregate of more than 267,000 sales transactions connected to roughly 22,000 actors in four countries over 20 years at daily resolution from 1801 to 1820. Striving towards a deeper understanding on multiple levels we disambiguate social dynamics of buying, brokering, and selling, while observing a general broadening of the market, where large collections are split into smaller lots. Temporally, we find annual market cycles that are shifted by country and obviously favor international exchange. Spatially, we differentiate near-monopolies from regions driven by competing sub-centers, while uncovering asymmetries of international market flux. Conceptually, we track dynamics of artist attribution that clearly behave like product categories in a very slow supermarket. Taken together, we introduce a number of meaningful network perspectives dealing with historical art auction data, beyond the analysis of social networks within a single market region. The results presented here have inspired a Linked Open Data conversion of the GPI, which is currently in process and will allow further analysis by a broad set of researchers. 2