Based on Art Sales from early in the eighteenth century to early in the twentiety century (mostly old master and early English pictures) by Algernon Graves, F.S.A. Published 1918-1921 in 3 Vols.
Art Sales provides a structured listing of art sales that happened in London, England, between the late 18th century (1791ish) to the early 20th century (1917ish).
The book lists results by artist name, then sales chronologically by year, month and date.
This is an extraction of that data in to CSV and JSON data formats. The data is being made available for curiosity and fun.
The data contained in Graves is a important for the study of provenance, but could also be used to do a larger network analysis on who was buying what, for how much, and from who.
The included data was initiated to track the purchasing activity of the Baring family, later to include the Earls of Northbrook, as part of a IMLS/NEH/Kress funded research project at Carnegie Museum of Art. Read more about that here, and check out our work on GitHub here.
What would be really useful is if this data could grow to include the entirity of the Art Sales book. You can find the first volume of the book online here. Vols. 2 & 3 are not currently online, but may be requested through a local library, likely through interlibrary loan- but check WorldCat first.~~
The first volume of Art Sales is available through the InternetArchive.
This data was obtained by scanning the second and third volumes of Art Sales in a book scanner with Optical Character Recognition (OCR) software. The first volume was already available in PDF format. The data was then extracted to xls format using ABBYY FineReader Pro, and then cleaned and exported to .csv and .json formats.
As of 10/28/2016, a great deal of data had been cleaned, but not all of it. If you are interested in seeing the complete data set, with some un-cleaned/unordered data, please see this Google Doc. The data on the first tab is a compilation of ALL of the tabs. It may be removed.
Because of Google's 2M cell limit on workbooks, a separate workbook has been started to contain the cleaned data. It is found here.
"artist": "Backhuysen, Ludolf", (who made the item)
"year": 1848, (year of auction)
"month_day": "June 2", (month and date of auction)
"auctioneer": "Christie's", "Paris" (place of auction or auctioneer)
"seller": "Sir Thomas Baring", (who is offering the item for sale)
"title": "Fresh Breeze", (title assigned to item)
"purchaser": "Nieuwenhuys", (purchaser of item, if recorded)
"pounds": 283, (price paid, pounds)
"shillings": 10, (price paid, shillings)
"pence": 0 (price paid, pence)
Check out the issues tickets for known issues with this data.
As of July 25, 2016, the biggest known issues are:
- The need to write scripts to extract
"seller"names from the front of artwork titles. - Troubles in modeling data where items were sold "en bloc". (i.e. a purchases bought paintings x, y, and z at the same auction on the same day for one price, rather than buying x, y, and z as separate units, with their own purchase prices.
Sure!
Send a pull request. Cheers.
Want to transcribe the rest of the data--not limited to the Barings? You. are. awesome. Clone this repo add the stuff, and send me a pull request. Enjoy endless good art history karma.
Anything you want. Nothing. Everything.
CC-0, and what have you.
It'd be nice if you ref'd this repo if you end up doing something cool with it, but not necessary. Data wants to be free.
The data contained in this file is related to the purchases of the Baring family. It's a much smaller pool of records, related to the art purchasing and selling activities of the Baring (Northbrook) families. You're welcome to use this data too, but it is much more limited in scope.
Use this data at your own risk. I've done by best to make it acurate, but I cannot swear or guarentee its acuracy. Please use your best judgement before basing a thesis on it.
This data set was scanned, cleaned, and harvested by Tracey Berg-Fulton. As Tracey is a human and had to do a lot of the data cleaning by hand w/software assistance, the data is likely not perfect.