How Art Auctions Work

On the evening of November 18, 2025, the auctioneer at Sotheby’s new Breuer building on Madison Avenue brought the hammer down on a Klimt at $205,000,000. The number the world read the next morning was $236.4M, the second-highest price ever paid for a work of art at auction. Both numbers are true. The gap between them, $31.4M, is the buyer’s premium, a fee the winning bidder pays the house on top of the bid, and it is the first thing you have to learn to see before an auction result means anything.

Six months later, on May 18, 2026, Christie’s sold Jackson Pollock’s Number 7A, 1948 from the S.I. Newhouse collection. The bidding stopped at $157M; the printed price was $181.2M. The Newhouse works alone made $631M that night, and the two back-to-back sales at Rockefeller Center totalled $1,121,126,500. All twelve of the works that sold above $40M in that May cycle, at Christie’s and at Sotheby’s, had been guaranteed before the sale by a third party who had agreed, for a fee, to buy the work if nobody else would. The catalogue told you so, in a small symbol beside the lot number that most readers never notice.

That is the shape of the whole market. Auction is the only place in art where a price is public, which is why every index, every record and every comparable a dealer quotes you is built from auction data. But the public price is manufactured: by a premium the buyer pays, a commission the seller may or may not pay, an estimate the house chose, a reserve nobody can see, bids the auctioneer is allowed to invent below that reserve, and a guarantor whose economics change what the number means. None of it is hidden. All of it is written in the conditions of sale at the back of the catalogue. This is the guide to reading it.

Every auction sale produces a hammer price, a buyer’s price and a seller’s price, and the three can differ by a third or more. The hammer price is the winning bid, the number the auctioneer repeats before the gavel falls. The buyer’s premium is a fee the winning bidder pays the house on top of the hammer, calculated on a published sliding scale. The seller’s commission is a fee the consignor, which is the trade’s word for the seller, pays the house out of the hammer, on terms that are private and negotiated. The house sits in the middle and is paid from both sides; the buyer and the seller never see the same number.

The number that reaches you is almost always the buyer’s price, because that is what the houses report. Christie’s and Sotheby’s publish sale totals and record prices “including buyer’s premium”, and the press repeats them. So when the Klimt is described as a $236.4M painting, what actually happened is that someone bid $205M and Sotheby’s added $31.4M in fees, a little over 15% of the hammer at that size, because the premium falls to 15% on the part of any bid above $8M. On a lot at the other end of the room the premium is 28%, so a $10,000 hammer becomes a $12,800 invoice before sales tax.

The record books, the indices and the “sold for” figures in a dealer’s pitch all carry the premium inside them. The academic return series mostly do too, which is one reason the fine-art guide on this hub, Investing in Fine Art, spends so long on the gap between the published return and the one an owner keeps. Investing in Emerging Artists, on the same hub, shows what the same premium does at the young end, where a first resale is the only exit and most works never get one.

The seller’s side is the part nobody publishes. The consignor of the Klimt did not receive $205M; the estate paid a commission, or negotiated one away, or was paid a share of the buyer’s premium on top of the hammer, and only the estate and Sotheby’s know which. On the biggest lots the premium is the only fee the house reliably earns, and everything in this guide follows from that arithmetic.

The premium is the one auction fee that is public, tiered and non-negotiable, and it has risen almost without interruption for fifty years. Christie’s and Sotheby’s introduced it in London in September 1975 at a flat 10%, citing inflation and rising costs after the 1973 oil shock. Dealers walked out of a Sotheby’s sale in protest, then kept bidding, and a 1981 High Court action by the dealers’ associations alleging collusion was withdrawn. The rate then climbed through the 1990s and 2000s in steps, 10% to 15% to 20% to 25%, each house matching the other within months. Two of those steps are verified: Sotheby’s was charging 15% on hammers up to $100,000 from January 1, 1993, and by 2023 the low tier stood at 26–27% at every major house (Antiques Trade Gazette, July 2023). The 20% and 25% steps between those dates are from our own reading of the trade press and were not re-verified for this piece.

The schedules as published at this guide’s as-of date, September 10, 2026, are these. Sotheby’s charges 28% of the hammer up to $2M, 22% on the portion from $2M to $8M and 15% above that, in every saleroom worldwide and every category except wine and spirits (a flat 24%), effective February 13, 2026. Christie’s adopted the identical 28%/22%/15% schedule at the identical $2M and $8M breaks (£1.5M and £6M in London) from September 1, 2026, in every department except wine, having charged 27% up to $1.5M for the twelve months before.

Phillips went higher, and earlier: 29% up to $1M, 22% to $6M and 15% above in New York from September 1, 2025, on a schedule it restated on April 12, 2026, with a “priority bidding” discount to 25%, 20% and 14% for a binding written bid at or above the low estimate lodged at least 48 hours before the sale. Bonhams, which works the middle market and reported $970M of sales in 2025 (house release, January 20, 2026), moves on October 1, 2026 to 30% on the first $35,000 of the hammer, 28% to $750,000, 25% to $1.5M, 21% to $7.5M and 14% above. Until then its scale, unchanged since 2023, runs 28% on the first $50,000, 27% to $1M, 21% to $6M and 14.5% above (Antiques Trade Gazette and ARTnews, September 2026).

Read the tiers as a marginal scale, like income tax. A $3M hammer at Sotheby’s pays 28% on the first $2M ($560,000) and 22% on the next $1M ($220,000), so the premium is $780,000, or 26% of the hammer. The effective rate falls as the hammer rises: about 22% on a $10M painting ($560,000 plus $1.32M plus $300,000), the full 28% on a $10,000 print, which is why the small buyer is the house’s best customer per dollar and the least courted.

The history since 2024 matters because it shows how quickly the schedule can move. On February 1, 2024 Sotheby’s announced an overhaul of its whole fee structure, in force from May 20, 2024: a buyer’s premium of 20% up to $6M and 10% above, the abolition of its 1% “overhead premium” on every lot, a new 2% success fee charged to sellers on hammers above the high estimate, and a standard seller’s commission of 10% on the first $500,000 of hammer with no commission on lots estimated above $5M (ARTnews, February 2024; The Art Newspaper, April 3, 2024). It lasted seven months before the reversal was announced.

In December 2024 Sotheby’s said the flat seller terms had “proved less attractive to potential sellers” and reverted, from February 17, 2025, to 27% up to $1M, 22% to $8M and 15% above, with bespoke terms for consignors; the overhead premium did not return and the success fee stayed (The Art Newspaper, December 19, 2024). Christie’s raised its tiers in September 2025, Sotheby’s went to 28% in February 2026, Christie’s matched in September 2026 and Bonhams follows in October: four increases in thirteen months, all in the same direction.

Two practical rules follow. The rate that applies is the one in force on the sale date, so check the conditions of sale printed in that sale’s catalogue. And the premium is charged on the hammer, but sales tax is charged on the hammer plus the premium; the invoice compounds.

A consignor’s terms are a private contract, and the published number is a starting point that the house abandons the moment a lot is worth fighting for. The nominal seller’s commission at the major houses is in the region of 10% of the hammer on a mid-value work, sometimes quoted at 15% for smaller lots, and it is charged on top of costs the consignor may also bear: cataloguing and photography, insurance while the work is in the house’s care, shipping to the saleroom, and sometimes a fee if the work fails to sell. Every one of those items is negotiable and the negotiation runs on one variable, which is how badly the house wants the work.

Sotheby’s publishes where the waiver starts, and it is higher than the trade usually says: the seller’s commission is waived outright on any consignment carrying a low estimate of $5 million or more. Between $20 million and $50 million the arrangement inverts and the published schedule says so — the consignor takes 40% of the buyer’s premium on top of the hammer, so the house is paying for the privilege of selling the work. The trade calls that an enhanced hammer. A consignor on an enhanced hammer receives more than the bid; the house’s only revenue on the lot is the part of the premium it kept. It is the economic centre of the evening sale: the trophy consignments that make a season are won by a package of enhanced hammer, guarantee, marketing budget and catalogue essay, and the house recovers the cost from the premium on the smaller lots around it, which is what the 28% pays for.

It is also why Sotheby’s 2024 experiment failed. A flat, published 10% seller’s commission was designed to end the haggling; consignors, who had spent thirty years learning that the published rate was a fiction, took their works to houses still willing to negotiate. The terms sheet is a menu, not a price list, and the house has more room than it admits, particularly if a rival is bidding for the same consignment.

The history has a darker chapter that explains why regulators watch the seller’s side. Between 1993 and 1999 Sotheby’s and Christie’s fixed the commission rates they charged sellers, a conspiracy that ran through both chairmen. The houses paid $512M to settle the civil suits in 2000; Sotheby’s pleaded guilty and was fined $45M; its chairman, A. Alfred Taubman, was sentenced to a year in prison and fined $7.5M, while Christie’s, which went to the government first, escaped a criminal fine under the leniency programme (U.S. Department of Justice, Antitrust Division, 2000–2002).

The two houses’ published premiums have moved in lockstep ever since, and the 2026 28%/22%/15% schedules are identical to the dollar. That is evidence of a duopoly and nothing more, but it is why the seller’s side, where terms are private, is where a consignor’s leverage lives.

The estimate is a marketing figure and the reserve is a contract term, and the two are linked by a rule that governs the whole sale. The estimate is the published range, low to high, printed beside every lot. The reserve is the confidential minimum below which the house will not sell, agreed between the consignor and the house before the sale. Sotheby’s guide for buyers states the rule that makes the pair readable: the reserve “is generally set at a percentage of the low estimate and will not exceed the low estimate of the lot”. A lot offered with no reserve at all is marked with a box symbol in the catalogue.

So the low estimate is a public ceiling on a private floor. If bidding stalls below the low estimate and the lot sells anyway, the reserve was lower; if the auctioneer passes the lot at a number just under the low estimate, you have found the reserve.

Whether estimates are honest is a question economists have been asking since Orley Ashenfelter’s 1989 paper in the Journal of Economic Perspectives, “How Auctions Work for Wine and Art”, which set out the mechanics of reserves, bought-ins and the auctioneer’s incentives. The sharpest answer came from Jianping Mei and Michael Moses in the Journal of Finance in 2005. Using thirty years of repeat sales, they found that pre-sale estimates for expensive paintings carried a persistent upward bias, and that works bought at high estimates went on to earn adverse abnormal returns. Their explanation is the house’s vested interest: a higher estimate wins the consignment and anchors the bidders, and the house is paid on the hammer either way.

A later study by Ekelund, Jackson and Tollison in the Southern Economic Journal (2013), built on more than 500 works by eight early twentieth-century American artists over twenty years and controlling for the lots that failed to sell, found the opposite sign: estimates biased downward, with the underestimate growing as the work’s value rose, which the authors put down to the house’s preference for telling a consignor the work made more than expected. The two findings sit together once you ask whose interest the estimate serves. At the top of the market, where the consignor is being courted, the estimate is a bid for the work; lower down, a modest estimate that the hammer beats flatters the house. In neither case is it a forecast of the price.

You can watch the mechanism in the marketing language. “Estimate on request” means the house does not want to print a number it may have to beat; a low estimate set conspicuously beneath the last comparable sale means the house wants the lot to appear to soar. The ratio of hammer to low estimate, the trade’s hammer ratio, is the statistic analysts use to judge a sale, and a house can engineer it by moving the estimate rather than the bidding.

Two more terms before the sale starts. A lot that fails to reach its reserve is bought in, and the trade treats it as damaged goods for years afterwards, because every future buyer can look up the failure. A lot withdrawn before it is offered does not count as unsold in the house’s statistics, which is why a marquee sale that “sold 98% of lots” may have quietly lost a few weak ones the night before. Neither convention is dishonest; both are reasons to read the lots offered against the original catalogue.

Below the reserve, the auctioneer is allowed to invent bids. The practice is called chandelier bidding, after the fixture the auctioneer appears to be taking bids from, and it exists because a sale cannot legally happen below the reserve anyway: the auctioneer is “bidding on behalf of the seller” to carry the lot up to the point where a real sale becomes possible, and to give the room the impression of competition on the way. Once the reserve is reached, only real bids count. That is the theory, and in most jurisdictions it is also the whole of the law.

New York City, which hosts the largest sales, used to regulate it. Under the city’s auction rules an auctioneer could take consecutive bids on the seller’s behalf only up to the reserve, and the practice had to be disclosed in the conditions of sale. Local Law 80 of 2021 repealed the city’s auction code, sections 20-278 to 20-290 of the administrative code, as part of a wider deregulation of licensed trades. The city withdrew its implementing auction rules with effect from April 10, 2022, and the repeal itself took effect on June 15, 2022 (Hunton Andrews Kurth and National Law Review client notes, 2022).

The Art Newspaper reported in May 2022 that the city had “removed rules governing auction houses in a bid to stimulate business”; the Center for Art Law and the Columbia Journal of Law and the Arts both noted that the specific prohibition on chandelier bids above the reserve, and the disclosure requirement, went with it. The houses say their conditions of sale still bind them to the old convention. The conditions are a private contract, and the reader should know that the rule they rely on is, since June 2022, the house’s promise rather than the city’s law.

The rest of the toolkit is less contested. The auctioneer sets the increments, typically about 10% of the current bid, and can split them to keep a hesitant bidder in. The book is the list of absentee bids the auctioneer executes for absent clients, and a lot opened well above its low estimate usually means the book already holds a bid there. The auctioneer also knows which lots are guaranteed and therefore cannot fail, and paces the sale accordingly. None of this is deception; it is choreography.

A guarantee converts an auction into a sale that has already happened, with the bidding left to decide the price. In a house guarantee the auction house promises the consignor a minimum, and if the lot fails to reach it the house pays and takes the work into inventory. In a third-party guarantee, which Sotheby’s calls an irrevocable bid, an outside party makes the same promise in exchange for a fee, and lodges a binding bid at the guaranteed level before the sale. If nobody bids higher, the guarantor owns the work at that price.

If someone does bid higher, the guarantor is paid: a financing fee for having taken the risk, plus, in most structures, a share of the amount by which the hammer exceeds the guarantee, typically 20% to 30% of that overage and reported anywhere from 10% to 50% (Center for Art Law, 2020; Boodle Hatfield). The houses differ on the fee. Christie’s pays its third-party guarantors a financing fee whether or not they end up buying the work; Sotheby’s has paid it only when the irrevocable bidder is outbid (Center for Art Law, 2020). The catalogue marks these lots with a symbol beside the lot number, and Sotheby’s conditions of sale state that the house and any guarantor “benefit financially if a guaranteed lot is sold successfully and may incur a loss if the sale is not successful”.

The mechanism was built for consignors, who want certainty, and it has ended up defining the top of the market. Bank of America and ArtTactic’s 2026 U.S. Art Market Report, published in March 2026, found that 78% of the value offered at New York evening sales in 2025 was guaranteed, the highest coverage of the decade, and that the houses were leaning on guarantees to win consignments more than ever before.

In one November Post-War and Contemporary evening-sale cycle across Christie’s, Sotheby’s and Phillips, ArtTactic counted 29 third-party-guaranteed lots out of 80, which was 36% of the lots and 56% of the value at the low estimate. The editorial that reports that count is undated; its own chart series begins in November 2015, so the season is a pre-2020 one, and we could not confirm which. The proportions stand.

In May 2026 all twelve works that sold above $40M, every one of them at Christie’s or Sotheby’s, carried a third-party guarantee (Artlyst, May 2026). Christie’s Newhouse sale on May 18 was 100% sold and, by The Art Newspaper’s account the next day, entirely backed by third-party guarantees; 30 of the 47 lots in the various-owner evening sale that followed were guaranteed by the house or a third party; and Constantin Brâncuși’s Danaïde drew a single bidder, its guarantor, and still set an artist record at $107.6M. The trophy market is, in a plain reading of those figures, a market in which the sale is arranged first and the auction is held afterwards.

What a guarantee does to the number is the part that matters for anyone using auction data. Suppose a work is guaranteed at $10M by a third party for a fee and a share of the upside, and the hammer falls at exactly $10M with no other bidder. The printed result is $10M plus premium; the “buyer” is the guarantor, who was paid a fee to bid, so the net price paid is below the print.

Suppose instead the hammer reaches $14M from a real bidder. The consignor receives $14M less any commission, but the guarantor takes a share of the $4M overage, so the seller’s net is lower than an unguaranteed $14M sale would have paid, and the guarantor, who may be an active collector of the artist, has just been paid to see the artist’s price rise. In both cases the result is real money and a real sale. In neither is it the clean meeting of a willing buyer and a willing seller that a comparable is supposed to represent.

The history shows what happens when the risk is mispriced. In 2007 Sotheby’s wrote $902M of guarantees. When the market turned in September 2008 the house lost about $52M on guarantees in the autumn season, by Apollo Magazine’s count; its third-quarter 10-Q had already booked $42M of guarantee losses against the fourth-quarter sales, and the full-year results of February 26, 2009 recorded “significant auction guarantee losses” in a quarter when net auction sales fell 46% (Sotheby’s Form 10-Q, Q3 2008; full-year 2008 results). Christie’s lost tens of millions on its own, a figure we could not re-verify, and both houses all but stopped writing house guarantees in the seasons that followed.

The market that grew back after 2010 moved the risk to third parties: hedge-fund managers, dealers and collectors who take the fee and the upside and post the capital. As of September 2026 it had not been tested by a fall on the 2008 scale.

The auction calendar is a hierarchy, and a work’s place in it tells you more about its price than the estimate does. The evening sale is a curated auction of a few dozen lots, held in the marquee weeks in New York (May and November), London (February, March, June and October) and Hong Kong, with the guarantees, the enhanced hammers, the catalogue essays and the press. The day sale follows, with hundreds of lots from the same departments at lower estimates, no marketing to speak of and far more bought-ins. The online-only sale is the third tier, a timed auction that runs for a week or two with no auctioneer and a closing time, and it is where the houses clear most of the prints, editions, photographs and lower-value works that used to fill the back of the day sale catalogue.

The figures show the concentration. The November 2025 New York season made about $2.2B across the houses, the strongest November in three years, Sotheby’s taking $1.17B of it and Christie’s $962.6M (Artnet News, November 25, 2025). The Lauder evening that opened Sotheby’s week sold every lot, and Sotheby’s evening sales for the week were reported at 98.5% sold by lot against 88.4% a year earlier (Artnet News and MyArtBroker, November 2025; a figure we could not re-verify against the house’s own results).

The May 2026 season went in with a combined low estimate of $1.8B and sold more than $2.5B: Christie’s about $1.4B with two day sales still to run, Sotheby’s $908.6M and Phillips $115.2M (The Art Newspaper, May 2026). The evenings carried it. Christie’s evening sales on May 18 and 20 contributed about $1.28B ($1,121,126,500 on the first night and $162.7M on the second), Sotheby’s Now & Contemporary and Modern evening sales $737M between them, and Phillips’s single sold-out evening $115.2M (ARTnews and The Art Newspaper, May 2026).

A handful of evenings, a few hundred lots, and the bulk of a year’s value. The Artprice count of the whole world, by contrast, was 1.28M lots offered in 2025, of which 867,000 sold; the evening sales are a rounding error in that number and most of the dollars.

Online has settled into a role rather than taken over. The Art Basel and UBS report for 2025, published March 12, 2026, put online art sales at $9.2B, down 11% on the year and the lowest since 2019, at 15% of the market’s value against a pandemic peak of 25% in 2020.

The high-value transactions moved back into the room; what stayed online was the volume and the new money: Christie’s said that 63% of its new buyers in 2025 made their first purchase online and that its average online buyer spent about $22,700 (ARTnews, December 2025).

For you that means two different markets under one brand: the online-only sale is where an outsider can buy a print by a blue-chip name for four figures, with the same 28% premium and the same conditions of sale, and the evening sale is where the numbers are made that the print will be valued against. The one art market that moved entirely online, NFTs, has its own guide on the NFT hub, Investing in NFTs.

Phillips’s priority-bidding discount, 25% instead of 29% for a binding written bid at or above the low estimate lodged at least 48 hours ahead, which cannot be cancelled and which the house introduced for the autumn 2025 season (Phillips, “Priority Bidding 101”), is the house buying certainty from buyers the way it buys it from guarantors. Artnet News argued in 2025 that it could fall flat, because a collector who commits early gives up the option to walk away when the room goes cold; that objection is also a fair description of a guarantee from the other side of the table.

A sale report contains five numbers, and the house chooses which to lead with. The total is premium-inclusive unless stated otherwise. The sell-through rate is the share of lots offered that sold, and it is computed after withdrawals, so a lot pulled the day before does not count as a failure. The hammer ratio is the total hammer divided by the total low estimate, a number above 1.0 meaning the sale beat its own floor. The share sold within or above estimate measures the estimates rather than the market. And the record, when one is claimed, is a premium-inclusive price compared against earlier premium-inclusive prices, in whatever currency the sale was held in, converted at the rate of the day.

The whole-market figures are less flattering than the marquee ones and more useful. Artprice’s report on 2025, published March 10, 2026, put the global unsold rate for fine art at auction at 32%, against 33% in 2024, across 1.28M lots offered and 867,000 sold, a 6.5% rise in lots sold on 814,300 the year before; 1,347 lots cleared $1M, up from 1,131. One lot in three offered at auction anywhere in the world found no buyer at the reserve, in a year the trade called a recovery. Against that, a marquee evening reported at 98.5% sold tells you that the evening sale is a different instrument, one in which the reserves are set to clear and the guarantees make sure they do.

Currency is the trap most readers fall into. London sales hammer in sterling, Hong Kong in Hong Kong dollars, Paris in euros, and the houses set their tiers in local currency at round numbers rather than at spot: Bonhams’s first tier from October 2026 is £25,000 in London and $35,000 in New York, and neither is a conversion of the other. A record “in pounds” can be beaten in dollars without any bidder paying more in either currency. Our own tape shows the problem at small scale: the Barnebys aggregation that feeds it recorded a 19th-century Rajasthani palace scene at $11,585.07 on September 8, 2026, a figure with cents because it was converted at a rate of the day. When you compare two results, convert both to one currency at the sale-date rate and say which one.

What a record measures, then, is this: the highest premium-inclusive price, in a stated currency, that a bidder was willing to pay, in a sale that may have been guaranteed, for a work whose estimate the house chose. It is a real number, and it is the number every index, valuation and insurance schedule will use. It is not, on its own, evidence that a second buyer exists at that level, and the artist’s day-sale results in the same week are usually the better guide to what you would get.

Most bids at a major sale arrive by telephone, by written instruction or through the house’s website, and the mechanics of each shape what you pay. An absentee bid is a written maximum lodged before the sale; the auctioneer executes it against the room at the lowest price that wins, so a $50,000 absentee bid against a room that stops at $30,000 buys the lot one increment above $30,000.

A telephone bid puts a house specialist on the line relaying the room to you and your bids to the auctioneer; you keep discretion but bid into a live sale with no time to think. Online bidding on the house’s platform does the same through a screen with a delay of a second or two. In every case the conditions of sale and the premium are identical to the room’s; third-party platforms that relay to smaller salerooms often add a percentage of their own.

Registration is where the house decides how seriously to take you; a new bidder on a marquee lot is asked for bank references or a deposit, because the house owes the consignor whether or not the winner pays.

Once the hammer falls you owe payment by the date in the conditions of sale, typically within days, and collection by a deadline after which storage is charged at your expense. Christie’s New York conditions, for one, allow storage charges on a lot not collected within seven days of the auction, and let the house charge interest, cancel the sale or resell if the total due is not paid by the payment date. The exact windows and tariffs vary by house and sale and change, so read them in the catalogue for the sale you are bidding in.

The condition report is the document that decides whether the price was a bargain. Every lot is sold as seen, and the house’s printed description is limited by the conditions of sale to matters of authorship, with a warranty period the two major houses’ published conditions set at five years from the sale for a work that proves to be a forgery (Sotheby’s Conditions of Business for Buyers; Christie’s Conditions of Sale), with far shorter windows for categories such as wine, gemstones and books. Condition is expressly excluded.

The house will supply a condition report on request, written by its specialist, and its own terms describe such reports as opinion rather than fact. So the report is a starting point. For any work you would mind losing money on, the sequence is to read the house’s report, view the lot in person or have an adviser do it, and commission an independent conservator’s report for a few hundred dollars before you set a ceiling. A relined canvas, an area of restoration under ultraviolet light or a print with a trimmed margin can cut a work’s value by a third, and none of it shows in the catalogue photograph.

The invoice is the hammer plus the premium, and the tax authorities treat the whole invoice as the price. In New York City, where the biggest sales take place, the combined state and city sales tax is 8.875%, and the buyer’s premium is part of the taxable receipt because it is part of the price of the goods (New York State Department of Taxation and Finance, advisory opinion TSB-A-24(48)S, October 2024; Sotheby’s guide for buyers). A $128,000 invoice on a $100,000 hammer therefore carries $11,360 of sales tax if the work is handed over in New York.

The exemption is delivery: a buyer outside the state who has the work shipped by the house’s shipping department or a common carrier to an address outside New York owes no New York tax. The buyer’s home state may impose a use tax, the charge a state levies at its own sales-tax rate on goods bought elsewhere and brought in, and several do. A buyer who collects the work in person, or sends a private van, pays. The rule turns entirely on where delivery takes place, which is why the shipping form is a tax document.

Shipping runs from a few hundred dollars for a framed work on paper sent domestically to several thousand for a large canvas moved by a fine-art shipper, plus brokerage, export licences and transit insurance for anything crossing a border. The flagship guide uses $2,000 for crating, shipping and framing on a $100,000 domestic purchase, and we keep it below.

Import duty into the United States is, for most fine art, zero. Original paintings, drawings and pastels are classified under heading 9701 of the Harmonized Tariff Schedule, original sculpture under 9703 and original prints under 9702, all with a general rate of “Free” (U.S. International Trade Commission, HTS 2026).

The 2025 tariff regime left that in place because of a clause older than the tariffs: the Berman Amendment, 50 U.S.C. § 1702(b)(3), which exempts “informational materials”, expressly including artworks, from actions under the International Emergency Economic Powers Act. Customs guidance in 2025 confirmed that headings 9701 to 9705 may qualify, subject to accurate documentation, and the trade press reported that fine art crossed the border untariffed while design objects, furniture and antiques more than a century old, which fall under heading 9706, did not (Artsy, Cultural Property News, Sullivan & Worcester, 2025; Renwick Fine Art Services, 2026).

Two cautions. Classification is decided at the border by Customs and Border Protection, and a shipment described loosely can be reclassified. And the exemption is for the tariff, not the tax: a work imported into the United Kingdom pays 5% import VAT, one of the lowest rates in Europe, and rates across the European Union vary by country; the flagship guide covers the freeport route that defers them.

The arithmetic of buying and selling one work through a saleroom is the whole argument of this guide, so here it is in dollars. Take a work hammered at $100,000 at a New York day sale in September 2026, at either major house. The buyer’s premium at 28% is $28,000, so the invoice is $128,000. Delivered in New York, sales tax at 8.875% on the invoice adds $11,360. Crating, shipping and framing, $2,000. Cash out on the day: $141,360, all of which counts as your tax basis, because the premium, the sales tax and the shipping are costs of acquisition (The Tax Adviser, “The taxation of collectibles”).

The same lot at Phillips would have cost $1,000 more in premium, or $3,000 less with a priority bid; at Bonhams from October 2026 it would cost $28,700 in premium; none of those choices moves the total by more than about 2%. Hold it for five years on a residential wall and insure it at the 1%-of-value end of the range the flagship guide cites, about $1,300 a year, or $6,500 over the hold. Total cash committed: $147,860.

Now consign it back to a saleroom and ask what the hammer has to be. Assume a 10% seller’s commission, no enhanced hammer (at this size you will not get one), and no unsold fee. Federal tax on a gain is at the 28% collectibles rate plus the 3.8% net investment income tax for a high earner, 31.8% in all (IRC § 1(h)(4) and § 1411; IRS Topic 409); state income tax is ignored here and would make every number worse.

The break-even hammer is about $167,700. At that price the net proceeds after commission are $150,900, the taxable gain over the $141,360 basis is $9,530, the federal tax is $3,030, and what you keep is $147,860, the cash you put in. The work has to rise 68% at the hammer in five years, about 10.9% a year, before you have your money back, on an asset whose honest long-run real return the fine-art guide puts at about 2.4% a year.

If the hammer instead comes back at $100,000, the price you paid, the round trip costs you $57,860: the premium, the sales tax, the shipping, the insurance and the commission, or 39% of the cash you committed and 58% of the hammer, with no gain to tax.

If the hammer doubles to $200,000, the net proceeds are $180,000, the gain is $38,640, the tax is $12,288, you keep $167,712, and the profit on $147,860 committed is $19,852 over five years, about 2.5% a year before state tax and inflation, for a work that doubled. The cost stack is the same one the flagship guide works over ten years; the shorter hold makes the fees loom larger against the gain, which is the point.

Three things in that table are within your control. Delivery outside New York removes the $11,360 if your home state does not claw it back in use tax. A priority bid at Phillips, or a house running a promotional premium on an online sale, trims the entry. And the seller’s commission is the one number on the exit side that a competing offer can move, even at $100,000, if the house wants the consignment for a themed sale. The premium is not negotiable, and it is the largest item.

Invest Alternative keeps its own record of the art market, read from public sources on a schedule, and it is narrower than the market and says so. The series and their dates are these, as of September 8, 2026. Artprice100, the blue-chip index of the hundred most-traded artists that Artprice publishes at year-end: our tape holds two observations, 91.7 at December 31, 2024 and 101.97 at December 31, 2025, the published levels, an 11.2% rise.

Masterworks activity, which we count daily from SEC EDGAR full-text search: 69 Form 1-U filings and 9 new offerings in the 90 days to September 8, 2026, up from 53 and 5 when the daily read began on September 1. Masterworks exits, a single read dated July 16, 2026: 8 exits in the trailing 400 days and an exit IRR of 11.5% on our own calculation, which is not the platform’s advertised figure. Masterworks and the fund route are the subject of the sister guide on this hub, Investing in Art Through Fractional Shares and Funds. And the IA Composite, our cross-asset index, at 100.271 on September 8, 2026, provisional, up 5.74% over 30 days and 0.29% over a year; art is one weight in it, not the whole.

The auction-specific part of the tape is the notable-sales feed, which we take from Barnebys, an aggregator of results across many salerooms, and which is useful precisely because it is not the evening sale. In the week to September 8, 2026 it recorded a Francis Newton Souza at $150,000 and a Krishnaji Howlaji Ara at $30,000, both on September 2, and the $11,585.07 palace scene discussed above on September 8. None carries a guarantee or a press release, and none tells you whether the premium is included, because aggregators report whatever the saleroom reports. That is the state of auction data below the marquee tier: available, useful for direction, inconsistent in what the number includes. When we quote our tape we say what it is and when it was read; it is never blended with Artprice or Art Basel and UBS.

The failures that cost auction buyers and sellers the most are structural, and each has a date. The first is the guarantee book turning against the house. Sotheby’s $902M of guarantees written in 2007 became a loss of about $52M in the autumn of 2008, by Apollo Magazine’s count, when bidders vanished and the house owned the works it had promised to sell; the consignors were paid, the shareholders were not, and the house all but stopped writing guarantees in the seasons that followed.

The third-party structure that replaced it moves that risk to guarantors who are not regulated, not disclosed by name, and often collectors of the same artists. A season in which the guarantors are the buyers of record on most of the top lots, which is what 78% coverage implies in a bad week, is a season in which the printed prices belong to a small group of people who were paid to bid.

The second is collusion, which the market has proved it can produce. The 1993–99 commission-fixing conspiracy described in section three ran across at least $400M of commissions charged to sellers in the United States, by the Department of Justice’s count, and left a structure in which the two houses still publish identical schedules. The safeguard is not the regulator, who acts after the fact, but the seller’s willingness to make the houses compete for every consignment, which is why this guide keeps returning to the written competing offer.

The third is attribution, which the auction contract handles by warranty rather than certainty. The most expensive work ever sold at auction, the Salvator Mundi that made $450.3M at Christie’s in November 2017, carries a Leonardo attribution many scholars dispute and has not been exhibited since; the fine-art guide tells that story and the Knoedler, Beltracchi and Philbrick cases beside it. The house warrants authorship for a limited period on limited terms, condition is excluded, and the remedy is a refund, not damages. Every lot is a bet on the catalogue entry, and the entry is the house’s opinion.

The fourth is the bought-in lot. A work that fails at auction is public knowledge, searchable on the price databases by anyone with a subscription, and the trade convention is that it should not return to the block for several years. The consignor who set a reserve too high to protect a price has, in a single evening, established a lower one. There is no fee for that in the terms sheet, and it is the largest cost on the seller’s side.

The fifth is the winner who does not pay. The house has already contracted with the consignor and, if it guaranteed the lot, already owes the money; a defaulting buyer leaves it to sue, re-offer or absorb the loss, which is why a house’s reported total can differ from the cash it collects. The best-known case is the Qianlong vase that hammered at £43M at Bainbridges, a small saleroom in west London, in November 2010, against an estimate of about £1M. With a 20% premium and VAT the bill passed £53M. The buyer never paid, and the vase was eventually sold privately through Bonhams in 2013, to a different buyer, for a reported sum of up to £25M (Bloomberg, December 2011; trade reports, 2013). The record stood in the press for two years. The consignors waited longer than that for less than half the printed price.

An outsider’s first auction purchase goes best when the sale is treated as the last step rather than the first. The sequence below is the one that loses the least money.

1. Read a year of results before you bid. Subscribe to the Artnet Price Database or Artprice, choose two or three artists you would actually want to own, and read every result for twelve months: the estimates, the hammers, the bought-ins and the guarantee symbols. You will learn the artist’s clearing price at the day-sale level, which is the price you will pay and the price you will get.
2. Register early and read the conditions of sale for that sale. The premium schedule, the payment deadline, the storage terms and the authenticity warranty are printed in the catalogue for the specific sale, not on a general page, and they change; the published schedules moved four times in the thirteen months from September 2025 to October 2026.
3. Get the condition report, then your own. Ask the house for its report, view the lot in person or through an adviser, and for anything above the low five figures pay a conservator for an independent report before the sale.
4. Set the all-in ceiling and convert it to a hammer. Divide by one plus the premium rate for that sale (1.28 at either major house in September 2026) and by one plus the sales-tax rate where the work is delivered (1.08875 in New York City), and write the resulting hammer down. That is your top bid.
5. Bid absentee or by phone, not in the room. Lodge the written maximum or hold to it on the telephone. The room is where written ceilings get raised.
6. Arrange delivery outside New York if you live outside New York, through the house’s shipper or a common carrier, and budget your home state’s use tax.
7. Keep the file. The invoice, the condition reports, the shipping papers and the catalogue page are the provenance you will sell with; the tax basis is the invoice plus the tax plus the shipping, and you will need every receipt in five or ten years.
8. When you sell, run the competition. Two written offers, the same eight questions to each, and a reserve set at your walk-away price rather than the house’s estimate.

The place to start is smaller than the place you are aiming for. A signed, numbered print by a blue-chip name in an online-only sale carries the same 28% premium, the same conditions and the same condition risk as a canvas at a hundred times the price, and teaches all of it for a four-figure ticket. What to buy there is the subject of the sister guide Investing in Prints and Editions.

The auction market publishes its own diagnostics, and the readings below are the ones that would change our view, each with its level as of September 10, 2026.

The premium ceiling

Bonhams’s move to 30% on the first $35,000 from October 1, 2026 is the highest low-tier rate any of the four houses has published, and Phillips already stands at 29%. If Sotheby’s or Christie’s follow above 28% on lots to $2M within a year, the round trip in the worked example is a floor, not a snapshot; if either cuts, the 2024 experiment is being retried and buyers should wait.

Guarantee coverage

78% of New York evening-sale value in 2025 was guaranteed. A reading above 80% in the Bank of America and ArtTactic report due in March 2027 would say the evening sale has become a placement business; a reading below 60% would say the guarantors have pulled back, which in 2008 preceded the prices doing the same.

The unsold rate

Artprice’s worldwide figure was 32% for 2025. A 2026 reading above 35% in the March 2027 report would mark stress below the marquee tier before it shows in the evening-sale totals; below 30% would be the strongest reading since the 2021–22 peak.

The November 2026 marquee week

May 2026 produced a $1.12B Christie’s evening and every $40M-plus lot guaranteed. A November week under $1.5B across the houses, or an evening sell-through below 90% at either major, would say the trophy supply of 2025–26 has been consumed.

Online share

15% of value in 2025, the lowest since 2019. A return above 20% in the 2027 Art Basel and UBS report would mean high-value lots are moving back to timed sales, where guarantees are rarer and the data cleaner.

Priority bidding

If either major house copies Phillips’s discount for binding early bids by the end of 2027, the premium has become negotiable for the first time since 1975, and every ceiling in this guide should be recomputed.

Our tape thresholds

The Artprice100 at 101.97 for 2025: a year-end 2026 print below 100 would put the blue-chip index under water for the cycle. Masterworks offerings at 9 per 90 days: a fall to fewer than 4 would say the fractional bid, which is a steady buyer at the day-sale level, has stepped back.