How Do Art Auctions Determine the Value of Artwork?

In May 2026, Jackson Pollock's Number 7A sold at Christie's in New York for $181.2 million. I had always assumed that a number like this emerges naturally in the saleroom — that competing bidders discover what a painting is worth in real time. What made me look closer is that auction results are the only prices the art market publishes. Galleries sell privately, but auction records become the market's public data: they set insurance values, loan collateral values, and the indices used to claim that art outperforms stocks. If that data is shaped before the bidding begins, then everything built on top of it inherits the distortion.

That sale made headlines around the world, but I realized the most important number was not the final hammer price—it was everything that happened before the auctioneer's gavel fell. Understanding those hidden mechanisms became far more interesting to me than the record itself.

How an Auction Price Is Constructed

Long before a lot is offered, a specialist assigns a low and high estimate based on comparables — prior results for similar works by the same artist, adjusted for size, period, condition, and provenance. The estimate is printed in the catalogue and functions as an anchor: it tells bidders what a reasonable price looks like. Separately, the house and the consignor agree on a reserve, the confidential minimum below which the work will not sell; it normally sits at or below the low estimate. During the sale, the auctioneer controls the speed and size of bidding increments and may open below the reserve, calling bids against the room until that floor is reached. If it is never reached, the lot is "bought in" — unsold. Finally, the buyer's premium, roughly twenty to twenty-six percent at the major houses, is added to the hammer price. The headline figure is therefore not one number but four decisions stacked together: estimate, reserve, bidding process, and fee.

Guarantees and the Pre-Sold Auction

The larger change is financial rather than theatrical. A guarantee means the house, or an outside party, promises the seller a minimum price regardless of what happens in the room. In a third-party guarantee — Sotheby's calls it an irrevocable bid — an outside guarantor commits to buy at an undisclosed level in exchange for a fee and a share of any upside, and may still bid on the lot. That guarantor knows the reserve and the guarantee; rival bidders do not. The practice now dominates the top of the market: in the May 2026 New York sales, all twelve lots that sold above $40 million, together worth roughly $851 million, carried third-party guarantees. Sell-through rates look spectacular as a result — Sotheby's Modern Evening Sale reached $303.9 million with 98 percent of lots sold — but observers described the bidding as measured and thin, and in the $630.8 million Newhouse sale several lots hammered below their low estimates. A high clearance rate can now measure how much risk was pre-sold, not how much demand exists.

For collectors, lenders, insurers, and investors, these published prices are more than headlines—they influence financial decisions far beyond the auction room. That makes transparency not only an art market issue but also a financial one.

The Problem and Possible Repairs

The result is a public auction that increasingly reports privately negotiated prices under conditions of unequal information. Three fixes would cost the industry little. First, post-sale disclosure: houses already mark guaranteed lots with a symbol, so they could also publish the guarantee amount and state whether the guarantor was the buyer. Second, published estimate methodology, listing the comparables used. Third, a shared, permissioned transaction and provenance registry — blockchain is well suited to this — combined with AI valuation models trained on historical results to produce an independent second opinion on estimates.

Rather than replacing specialists, AI could serve as an independent benchmark by identifying estimates that deviate significantly from historical pricing patterns. Likewise, blockchain would improve confidence only if market participants choose to record complete and reliable transaction data.

Technology alone is not enough: a ledger can only certify what participants choose to record.

Conclusion

An auction is less a machine that discovers value than a stage where a privately arranged value is performed in public and then converted into data. Guarantees are not fraud; they bring great works to market and underwrote the market's 2026 recovery. But if art is to function as a serious asset class, the prices it publishes have to mean what investors think they mean. The cost of disclosure is one evening of mystique. The benefit is a price record worth trusting.

The more I studied auction markets, the more I realized that understanding art requires understanding finance as well. The future of the art market will depend not only on extraordinary works of art, but also on transparent systems that allow those works to be valued with confidence.