Reading an Auction Estimate

What the numbers are, and what sits behind them

A lot carries a low and a high estimate. Both exclude the buyer's premium, which is defined by every house as a percentage added to the hammer price. Behind the published pair sits the reserve, the confidential minimum below which the house will not sell. The reserve cannot exceed the low estimate, which means the low estimate is a soft floor carrying real information: whatever the reserve is, it is at or under that number.

An estimate is a commercial forecast rather than a valuation. It is set to attract bidding, which frequently means pitching low enough to start a contest and to make the lot look like an opportunity. It is also the only part of a catalogue the department has to defend in public within a few weeks, in a room, against people who trade the same material.

That accountability is what makes it worth reading first. Catalogue prose is written to sell and carries no cost when it overstates. An estimate that is wrong is visible to everyone in the saleroom on the day, and it stays on the record afterwards as a measurement of what the department believed.

The estimate is also the product of a negotiation the reader never sees. A consignor with several houses competing for a collection is shown proposed ranges, and the range that wins the consignment is not always the range the specialists would have set on their own. A conspicuously optimistic estimate on an otherwise unremarkable lot is frequently a residue of that competition rather than a judgement about the object, and it tends to appear on the lots that anchor a sale rather than on those that fill it.

What the spread tells you

A tight spread indicates confidence: comparable material has traded recently at a predictable level and the department expects the lot to land in a narrow band. A wide spread indicates that the outcome cannot be predicted, which usually reflects thin comparables rather than doubt about the object itself. Unique material has no recent trades to anchor it, so the range widens to cover the plausible outcomes.

The extreme case is the unpublished estimate. 'Estimate on request' or 'refer department' signals a figure large enough that publishing it would either deter bidders or embarrass the seller if it went unmet. It is a commercial device rather than a scholarly one, and its presence tells a reader about the consignment negotiation rather than about the picture.

A spread should also be read against the material around it. A single lot with an unusually wide range in a sale of tightly estimated comparables is the department admitting that this object does not belong to the same market as its neighbours, which is worth investigating whatever the catalogue essay says about it. The reverse configuration, a narrow range on an object with no obvious comparables, is a claim that the department knows something the reader does not, and it is occasionally simply an error.

The spread should also be read against the heading. A confident, unqualified attribution over a diffident estimate indicates that the department is less certain than its own wording, since the estimate is the number specialists must justify internally and the heading is the number they must justify only if challenged within the warranty period.

When a result beats an estimate, and when it only appears to

Because estimates exclude the premium and published results include it, a lot can be reported as exceeding its estimate while having sold at or below it. This is the most common numerical distortion in art reporting and it consistently overstates market strength.

The Frans Post view of Olinda at Sotheby's New York on 21 May 2025 is the plain case. The estimate was USD 6 million to 8 million; the hammer was USD 6 million, the low estimate exactly; the widely reported figure was USD 7.3 million with fees. A result at the bottom of the range was published as a number in the upper half of it, and both figures were accurate.

Turner's Rising Squall at Sotheby's London on 2 July 2025 shows the same effect at a different scale. The estimate was GBP 200,000 to 300,000, the hammer GBP 1.5 million and the total GBP 1.9 million with fees. Coverage described the result as nearly eight times the estimate, which is the premium-inclusive total measured against the low estimate. Hammer against high estimate is five times, which is still a remarkable result and is the correct comparison. The right test is always hammer against estimate; anything else compares two different quantities.

The estimate as a record of what was known on the day

An estimate is also a historical document about the state of knowledge at the moment of the sale, and it is one of the very few numbers in this market that measures research directly. The Cimabue Mocking of Christ, found during a house clearance near Compiegne and sold by Acteon at Senlis on 27 October 2019, was estimated at EUR 4 million to 6 million after the attribution work by Eric Turquin had been done. It hammered at EUR 19.5 million, with a premium-inclusive total of EUR 24.18 million, contested by seven registered bidders of whom three were in the room.

What the estimate does not capture is the state of knowledge before that work began. The auctioneer Philomene Wolf, who found the panel above a hotplate during the clearance, recorded that the family considered it special but thought it was an icon. A figure of EUR 300,000 to 400,000 is often quoted as the pre-attribution appraisal; it does not appear in the primary reporting of the sale and should not be relied on without a source.

The Camden portrait measures the same interval from the other side. Thomaston Place estimated it at USD 10,000 to 15,000 as a work after Rembrandt, and it hammered at USD 1,175,000, roughly eighty times the high estimate. That number is a measure of what the room knew that the catalogue did not, and it is the honest arithmetic of what attribution research is worth.

The estimate is where the department's real opinion lives

Read the estimate before the essay. The prose describes, praises and hedges at no cost. The estimate has to survive contact with a saleroom, and it is set by people who will be asked afterwards why it was wrong. Where the two disagree, the estimate is the more reliable signal of what the specialists actually think.

The relationship to the warranty completes the picture. A house's authenticity guarantee attaches only to the unqualified heading, so a picture catalogued at a lower rung is sold with the house expressly declining responsibility for authorship. A low estimate under a qualified heading is therefore two forms of the same caution stated twice: the department will not assert the attribution, and it will not price the lot as though the attribution held.

Where those two signals diverge, something interesting is happening. A qualified heading over a high estimate means the house expects the market to disbelieve its own caution, which is a common configuration in old master sales and is precisely the situation in which a specialist buyer looks harder.

Bought in, and what a failure records

A lot that fails to reach its reserve is bought in and remains the consignor's property. It has not sold, and its estimate stays on the public record as a claim that the market declined. Buy-in rates across a sale are one of the few honest indicators of demand, and they are reported far less often than totals.

An unsold lot also carries a practical cost that rarely reaches print. A picture that has been offered publicly and failed is described in the trade as burned, and it commonly has to rest for a period before being reoffered, because the failed estimate is now a matter of record that any bidder can look up. The estimate that was set to attract bidding has become a documented ceiling.

This is the strongest argument for reading estimates as evidence rather than as marketing. They are set with commercial intent, defended in public, and preserved whether they succeed or fail. Very little else in a catalogue is subject to all three conditions.

For a record-keeper the practical rule follows from that. Store the estimate range alongside the result, in the currency of the sale, and note whether the lot sold or was bought in. An entry giving only the price achieved has discarded the department's own prediction, which is the one piece of the catalogue that was tested. Over a run of sales those stored estimates become a usable measure of how well a market was understood at a given moment, and they are the only measure of it that anybody published in advance.