London’s Auction Estimate Paradox: Why Higher-Estimated Lots Sold More Often in CollectWisely’s 2025 Sample

The upper estimate is one of the most visible numbers in an auction catalogue. It looks like a price forecast, a confidence signal and, sometimes, a verdict on quality. Our London data shows why it is dangerous to treat it as all three.

The global art market returned to growth in 2025. Arts Economics estimated sales of $59.6 billion, up 4%, while public-auction sales increased 9% to $20.7 billion. The recovery was led by higher-value work, and the UK retained an 18% share of global sales.¹ London sat at the centre of that rebound—but its headline statistics hide two very different markets.

CollectWisely analysed every qualifying GBP-denominated fine-art record in the CollectWisely Price Database dated from 1 January to 31 December 2025: 22,029 lots, 23 auction houses and 274 house–auction-number–date groups. We counted a lot as sold when the record contained a positive reported price. Published estimates were available for 99.6% of the sample. This is not an official census of every London lot.

The headline is true—and incomplete

At first glance, the pattern could not be cleaner. Lots with an upper estimate below £5,000 sold 74.7% of the time. Sell-through then rose in every band, reaching 86.3% between £250,000 and £1 million and 90.4% above £1 million.

In the recorded sample, the £1 million-plus estimate tier had a 15.7 percentage-point higher observed sell-through rate than the sub-£5,000 tier. That does not mean the price tag caused the sale.

The market’s volume and value live in different places

The first clue lies in concentration. Among the 21,943 lots with usable estimates, lots estimated below £50,000 made up 91.3% of lots and 12.5% of reported value; the £1 million-plus tier represented 0.6% of lots and 48.5% of that estimated-lot value. Across the complete 22,029-lot sample, the top 1% of sold lots—167 works—accounted for 57.4% of reported value.

The year’s largest London result makes the imbalance tangible. Canaletto’s Venice, the Return of the Bucintoro on Ascension Day sold at Christie’s for £31.935 million, equal to 4.1% of the entire sample’s reported value on its own. Christie’s had marketed the estimate on request, indicating expectations above £20 million.^{2 3}

Three more works illustrate the range at the top: Francis Bacon’s Portrait of a Dwarf realised £13.11 million against a £6–9 million estimate; René Magritte’s La reconnaissance infinie realised £10.315 million against £6–9 million; and Yoshitomo Nara’s Cosmic Eyes (in the Milky Lake) realised £9.028 million against £6–8 million. These were not merely expensive objects drawn at random from the same pool as a £1,500 print. They were selected, researched, marketed and placed in globally visible sales.

The reversal: compare like with like

Four international houses—Christie’s, Sotheby’s, Phillips and Bonhams—accounted for 45% of estimated lots but 135 of the 136 lots in the £1 million-plus tier. Their aggregate sell-through improved with every price band. At London specialist and regional houses, the opposite pattern appeared across the well-populated bands: 74.2% below £5,000, 66.9% at £5,000–£20,000, 56.0% at £20,000–£50,000 and 47.4% at £50,000–£250,000.

We then estimated a logistic model of sale probability using the logarithm of the upper estimate, controlling for auction house, fine-art category and sale month, with standard errors clustered by house–auction-number–date group. In the raw model, a tenfold increase in estimate had no statistically reliable relationship with sale odds (odds ratio 1.073; 95% CI 0.949–1.213). After controls, the direction reversed: a tenfold higher estimate was associated with 23% lower odds of sale (odds ratio 0.772; 95% CI 0.694–0.858; p<0.001).

The finding held when Christie’s was removed (odds ratio 0.774), and separately for paintings and works on paper (0.799) and prints and multiples (0.727). For the four international houses alone, the adjusted estimate effect was smaller and statistically inconclusive (0.897; 95% CI 0.777–1.037). Across each of the seven largest houses, the separate unadjusted estimate slope was non-positive. In other words, the high-end advantage exists between market segments, not consistently within them.

Why high-end art can clear more reliably

1. The expensive catalogue is pre-filtered

High-value evening sales are shortlists, not inventories. Houses compete for consignments, decline work that cannot support its positioning and concentrate specialist attention on a small group of lots. In April 2025, Bonhams reported that a 15-lot London evening sale sold 93% by lot and 99% by value.⁴ Phillips’ March London evening sale sold 26 of 29 lots, or 90%. Its own release described the strategy as curating focused sales.⁵ Those figures are house-reported and promotional, but they illustrate the selection process visible in our broader data.

2. Reserve prices create a discontinuity

Auction outcomes are not governed by the estimate alone. Christie’s explains that a confidential reserve is agreed with the seller and cannot exceed the low estimate; if bidding does not reach it, the lot is bought in.⁶ Economic research has found reserves commonly sit around 70% of the low estimate, making “sold” a threshold outcome rather than a smooth reflection of demand.⁷ Two similarly estimated works can therefore have different sale probabilities because of different reserves and seller constraints.

3. Financial engineering is concentrated at the top

Guarantees, irrevocable bids and third-party backing can reduce the risk that a trophy lot fails publicly. Christie’s describes minimum-price guarantees and third-party guarantees as financial arrangements that may secure a sale while transferring some risk.⁸ Sotheby’s disclosed both a guarantee and irrevocable bids for Bacon’s Portrait of a Dwarf, which sold above its high estimate.⁹ Academic evidence likewise finds that more expensive lots are more likely to carry guarantees, although guarantees do not automatically raise prices after controlling for value.¹⁰

4. Global demand meets scarce, recognisable supply

The 2025 market recovery was strongest at the high end. That matters because a small number of established artists and canonical works can draw bidders across geographies even while demand below the trophy tier remains fragmented. The £1 million-plus segment is therefore both scarce and unusually exposed to global marketing, private-client networks and pre-sale interest.

Category still matters

Price was not the only dividing line. Paintings and works on paper—the largest category, with 12,427 lots—sold through at 76.0%. Prints and multiples reached 75.5%, and sculpture and three-dimensional art 77.0%. Photography and video trailed at 68.2% across 1,367 lots, a gap too large to ignore when evaluating an artist, house or estimate strategy.

What the paradox means in practice

For collectors

A high estimate is not, by itself, evidence of deeper demand or lower risk. Compare the work with lots in the same house, category and sale format. Check the reserve framework and financial-interest symbols. Most importantly, include unsold lots in comparables; a sold-only history systematically removes the observations where price expectations failed.

For sellers

Estimate discipline matters, but placement may matter more than a simplistic “price it low” rule. A tightly curated evening sale with active pre-sale interest is structurally different from a broad multi-category auction. Ask how the house arrived at the estimate, how many close comparables sold and failed, what reserve will be agreed, and whether the house expects financial arrangements or irrevocable bidding.

For researchers and journalists

Do not infer market health from sold prices alone. Unsold works have no observable transaction price, creating a selection problem documented in auction research.¹¹ Pre-sale estimates can themselves be biased, and analyses that ignore the no-sale mechanism risk overstating returns or misreading demand.¹² Sell-through should be reported beside price performance, not as an afterthought.

London’s premium tier did not sell more often because expense made it safe. It sold more often because the objects, houses, sale formats and financial structures were different before the first bid was placed.

Methodology and limitations