The Price Is in a Currency That No Longer Exists

A price is a number and a unit, and the unit is usually gone

A recorded price has two halves, and the trade quotes only the first of them. Across the 187 auction results in the published records of this archive, 106 state what money the transaction was in, and those 106 rows use sixteen different descriptions of it.

Some of those descriptions are the same currency written twice, which is itself a finding: the corpus contains Dutch guilders and NLG, French francs and FRF, and four separate ways of saying pounds sterling because two of them specify that the bidding was in guineas. Collapsed to actual units of account, the corpus is holding about ten, and only four of them are money a reader can convert today.

Every league table of the most expensive works ever sold is a conversion, and the conversion is the argument rather than the data. The archive's own rule is to record the currency of the sale and to treat any conversion as derived and date-dependent, which is why the rows above exist at all.

The guinea stopped being a coin in 1816 and went on setting records until 1970

The guinea was fixed at twenty-one shillings from 1717, ceased regular production in 1813 and was demonetised in the Great Recoinage of 1816, when the gold sovereign replaced it. It then continued for more than a century and a half as a unit of account in which professional fees, horses, land and works of art were quoted, surviving decimalisation in 1971 by a couple of years.

The consequence for this archive is a picture with two guinea prices 169 years apart. Velazquez's portrait of Juan de Pareja was sold at Christie's in 1801, catalogued as a portrait of a Moorish slave who became a great painter, for 39 guineas. It went to the Radnor family and stayed there. On 27 November 1970 Christie's sold it again, and the bidding was taken in guineas: 2,200,000 guineas, which is 2,310,000 pounds, the first painting to pass a million pounds at auction and the world record price for any painting at the time. Alec Wildenstein bought it for the Metropolitan Museum of Art.

Three things follow that a price series will not tell you. The 1801 figure and the 1970 figure are in the same unit, so they are genuinely comparable, which is rare in this corpus and is an accident of British saleroom conservatism. No buyer's premium was charged on either, because the modern premium was introduced in London in September 1975. And the dollar figure of 5,544,000 that circulated in 1970 is a conversion made at the exchange rate of that week, which is a fact about sterling in November 1970 and not a fact about the painting.

Eight guineas, then five, at the same auction house eleven months apart

The guinea is also the unit in which this archive records the two cheapest sales of a major Caravaggio. The National Gallery of Ireland's own catalogue entry for The Taking of Christ gives the sales in its provenance: Dowell's, Edinburgh, 25 June 1921, lot 30, to John Kent Richardson for eight guineas; and Dowell's again, 3 May 1922, lot 480, to Major Charles Noel for five guineas.

Eight guineas is eight pounds and eight shillings, or 8.40 pounds decimal. Five guineas is 5.25 pounds. The picture lost three guineas in eleven months in the same room under the same wrong attribution, and the interest of the figure is not the smallness of it but the unit: an object catalogued as a seventeenth-century Dutch painting was priced in a coin that had been out of circulation for a century, because that is what the saleroom quoted in.

The same entry carries the commission price. The gallery records that Caravaggio made the picture for Marchese Ciriaco Mattei in 1602 for 125 scudi. Between the scudo of 1602 and the guinea of 1922 there is no exchange rate that any source in this record supports, and the two numbers cannot be put in the same column.

The same object priced in florins and then in francs

The Harvesters carries two recorded prices and they are 222 years and two currencies apart. The Metropolitan's published provenance gives the first as a sale by the Antwerp dealer Hane von Wijke in 1594, for 1,400 florins, to the Antwerp City Council, which bought the picture as a gift for Archduke Ernst, governor of the Netherlands.

The panel then went through the imperial collections at Prague and Vienna for two centuries, left them in 1809 with the comte Antoine-Francois Andreossy, and reappeared at the Perignon sale in Paris on 11 March 1816 as lot 1. It was catalogued as the work of a master of the old German school, a landscape of the greatest detail showing the labours of the harvest, measured in pouces, and dated 1546. It made 90.11 francs: ninety francs and eleven centimes.

The second figure is not a price for a Bruegel. It is a price for an anonymous German landscape misdated by nineteen years, and the currency change makes that hard to see. A conversion between 1,400 florins in 1594 and ninety francs in 1816 would produce a number, and the number would be describing a catalogue error rather than a market.

One object, three units, and a revaluation in the middle

The Fortune-Teller has never been sold at public auction, and every figure attached to it comes from an inventory or a private treaty. That makes it the clearest case in the archive of what happens when a single object accumulates valuations in a currency that keeps being replaced.

The Metropolitan's provenance records a posthumous inventory of Jean-Baptiste de Bretagne in 1650, number 14, in which a large canvas of fortune-tellers described as an original by La Tour is valued at thirty livres, one of the highest valuations in an inventory that contained five La Tours. It records an 1879 inventory in which a large signed La Tour of the Bonaventure is valued at 250 francs. And it records the sale by the heirs of General de Gastines in 1949 to Wildenstein for 7,500,000 francs.

The trap is in the last figure. The 1949 sum is in old francs. On 1 January 1960 France introduced the new franc at a rate of one hundred old francs to one new one, so 7,500,000 francs of 1949 became 75,000 francs of 1960 without anything happening to the picture. A price series that carries the 1949 figure forward unadjusted overstates it by two orders of magnitude, and there is nothing in the number itself to warn you.

Sometimes the consideration is not money

The most extreme entry in this archive's currency field describes a payment that was not made in any currency. In the second volume of Ruins of Desert Cathay, published in 1912, Aurel Stein set down what he handed over at the Mogao caves in 1907 in his own words: for all these acquisitions four horse-shoes of silver, equal to about Rs.500, passed into the priest's hands; and when I surveyed the archaeological value of all I could carry away for this sum, I had good reason to claim it a bargain.

Two pages earlier he gives the rate he was working to. He records hinting at forty horse-shoes, about 5,000 rupees, and being resolved to give twice as much if he had to, whatever the excess over his sanctioned grant. Forty ingots to five thousand rupees puts one ingot at about 125 rupees, which makes his four-ingot figure internally consistent.

Three qualifications belong beside that number and this archive states all three. The sum bought fifty bundles of Chinese text rolls and five of Tibetan ones as well as everything Stein had already selected, so it is not a price for the Diamond Sutra scroll. The rupee valuation is Stein's own, made by the buyer. And no source consulted for the record states a sterling equivalent at the time. What is in the field is a weight of metal and a buyer's estimate of it, and treating that as a price is an act of interpretation, not of reading.

Where the unit survives but the arithmetic does not

Pre-decimal sterling is the quiet version of the same problem, because the currency did not die and the arithmetic did. This archive records the Fonthill vase selling at Christie, Manson and Woods in June 1882, in the Hamilton Palace dispersal, for 28 pounds and 7 shillings, which the record also states as 28.35 pounds because a shilling was a twentieth of a pound.

Whistler's Princesse du pays de la porcelaine sold at Christie's on 28 May 1892, lot 39 of the Leyland estate sale, for 420 guineas. The Portland Vase made 980 guineas at the sale of the Duchess Dowager of Portland's museum on 7 June 1786, lot 4155, bought by an agent for her son, and she had paid 1,800 guineas for it and three other antiquities two years earlier. Velazquez's Pareja made 40 pounds 19 shillings in 1801 according to the Metropolitan and 39 guineas according to Buchanan's account of 1824, which are the same sum written in two units.

None of these figures is wrong and none of them is difficult, but each of them requires the reader to know a conversion that is no longer taught. The error a modern table makes with pre-decimal sterling is not a failure of exchange rates; it is reading 28 pounds 7 shillings as 28.07.

What to do with a price whose currency is gone

The practical rule is to keep the number and the unit joined, and to refuse to make the conversion inside the record. This archive records the currency of every sale as a field of its own precisely because a hammer figure detached from its unit is not evidence of anything.

Three habits follow. Quote the figure in the words of the document, including the shillings, the centimes and the guineas, because the unit tells you which market you are in and often which decade the source was written in. State who made any conversion and when, as with Stein's own rupee valuation of his silver, since a conversion by the buyer is testimony rather than data. And when two prices for the same object are in different units, say so instead of interpolating, because the interpolation will look like a measurement to everybody who reads it afterwards.

The reason this matters beyond bookkeeping is that price series are used as evidence about objects. A repeat sale, a market index and a most-expensive-ever ranking are all built by putting figures in a column, and the column is only meaningful if the figures were in the same money. In this corpus they usually were not.