Investing in Signed and Vintage Jewellery

In late April 2020, with the world’s salerooms shut, Sotheby’s ran a single-lot online sale in which nobody handled the piece, nobody sat in a room, and the auctioneer was a countdown clock. A Cartier Tutti Frutti bracelet made around 1930 — carved Indian rubies, sapphires and emeralds set in platinum in the Art Deco manner Cartier had been building since the 1920s — sold on April 28 for $1.34 million against a $600,000 to $800,000 estimate, pursued by five bidders and setting the record for a jewel sold in an online auction. The stones in it were not, by the standards of the top of the gem market, extraordinary. Carved coloured stones of that kind are bought by the gram in Jaipur. What the buyer paid for was the fact that a particular house had assembled them in a particular decade, and that the trade agreed on which house and which decade.

That is the whole thesis of this guide, and the four years since have sharpened it rather than softened it. Over the same stretch the price of the thing most people mean when they say “jewellery” — a white diamond in a commercial size — fell by about half, because a laboratory in Gujarat learned to grow the identical object for a tenth of the money. The stone market broke. The signature market did not.

There is an irony in it that the trade has been slow to say out loud. For a hundred years the diamond was sold as the store of value and the setting was sold as decoration. The arithmetic has inverted. In 2026 the durable, non-reproducible, hard-to-fake part of a fine jewel is the name inside the clasp, the archive entry behind it and, increasingly, the weight of gold holding the whole thing together. This guide is about how that market works, what it honestly returns, what a round trip costs, and how somebody outside the trade buys into it without being the exit liquidity. The hub’s flagship, Investing in Luxury Goods, covers the whole boutique economy from handbags down. Investing in Luxury Watches covers the parallel market in signed objects on a wrist, and Investing in Sneakers runs the same test at the other end of the price scale, where the premium is set by a release date rather than by a period. This one stays in the jewellery case.

A jewel can be taken apart. That single physical fact governs the entire market, because it means every piece has a floor and a ceiling that can be calculated independently of taste. Melt the metal and you get a number from the London gold fix. Unset the stones, send them for grading, and you get a number from a dealer’s price sheet. Whatever the piece fetches above the sum of those two numbers is what the market is paying for design, workmanship, condition, provenance and the name — and only the name is genuinely scarce, because the metal is a commodity, the stones now have a manufactured substitute, and the workmanship can be copied by any good bench in Valenza or Bangkok.

Call the difference the signature premium. On our reading of resale asking prices against retail across current production, a bracelet stamped Cartier, Van Cleef & Arpels, Bulgari or Tiffany carries a retail premium of roughly two to four times the intrinsic value of its metal and stones, and about half of that premium survives into the resale market. That is the number to hold in your head, and it does two things at once. It tells you why signed jewellery is worth studying: no other component of the piece has held any premium at all. It also tells you why buying at retail is a losing trade: you pay for the whole premium at the counter and recover half of it, which is a 30% to 50% haircut on day one before any fee.

The mechanism separating good outcomes from bad ones is not the house but the period. Current production is made to order in quantity — a Love bracelet is on the website, nobody is rationed, and the resale discount to retail narrows with each price rise without ever closing. Discontinued production from a defined era is finite, and its supply only shrinks as pieces are broken up for their stones. This is the same mechanism that governs vintage watches, and if you have read Investing in Luxury Watches you already know the shape of it: signature, period, condition, box and archive letter do the work, and the movement — here, the stones — is the smaller part of the price. What jewellery does not have is the rationing that drives the other two luxury markets on this hub. Nobody waits for a Love bracelet the way buyers wait for a Birkin or a steel Daytona, so the allocation premium that Investing in Hermès and Investing in Rolex are largely about simply does not exist in the jewellery case, and the only premium left is the one the period earns.

One bidder underwrites that premium, and it is the maison itself. Cartier has run a commercial vintage arm, Cartier Tradition, since 1996; Van Cleef & Arpels began its Heritage collection in 2007 and keeps a Patrimony department behind it with the client files, account books and sketch libraries. Both buy their own historic production back — usually from clients, sometimes from dealers — restore it, and then either exhibit it or sell it again. That last clause is the one that matters to a buyer, and it is the one the trade usually leaves out.

A house that only collected would be a museum quietly withdrawing supply. A house that restores and resells is a trade buyer with better information than anybody else in the room, a balance sheet that does not need the piece to clear this season, and a commercial reason to pay up. That cuts both ways. It puts a well-informed underbid beneath the best-documented pieces from a defined period, which is the strongest support the signature premium has and the reason it has held while the stones and the workmanship have not. It also means that on exactly those pieces — the ones this guide tells you to want — the person you are bidding against may be the company whose name is stamped inside them. Section three has the public proof of it, and section nine has the mechanics.

Four names carry most of the liquidity, and they carry it because each ran a period in which it produced a recognisable, documented and now finite body of work. Cartier is the deepest market: the Art Deco output of the 1920s and 1930s, the carved-stone Tutti Frutti pieces, the Panthère jewels of the era in which Jeanne Toussaint ran fine jewellery from 1933, and the mid-century Trinity and Love designs. Van Cleef & Arpels trades on invention rather than volume — the Mystery Setting patented in 1933, the Zip necklace of the 1950s, the Alhambra motif from 1968. Bulgari is the Roman house whose Serpenti, first made as a watch in 1948 and given its coiled tubogas form through the 1960s and 1970s, is the most distinctive object of the postwar period. Tiffany & Co. trades on its designers rather than the house: Jean Schlumberger, who began designing for Tiffany in 1956 and was the first designer the house allowed to sign his work, and Elsa Peretti and Paloma Picasso after him.

Below the four sit the independents, and they are where the highest per-piece prices in the contemporary market get made. JAR — Joel Arthur Rosenthal, working from 7 Place Vendôme in Paris since 1977 — makes seventy to eighty pieces a year, sells to a closed list of clients with no shop window and no advertising, and his work routinely clears its high estimate at auction by multiples; the collection assembled by the actress Ellen Barkin made $20.4 million with premium at Christie’s New York on October 10, 2006, the largest group of JAR to reach the market at once. That shape — tiny output, a closed client list, and secondary prices set by the people who cannot get on it — is the one Investing in Independent Watchmakers describes on a wrist.

Suzanne Belperron, working from the 1930s to the 1970s, is the instructive counter-example: she refused to sign her work on the grounds that her style was her signature, and the market has since had to authenticate her pieces by archive drawings and by attribution rather than by a stamp. The cost of that is measurable: of the sixteen Belperron lots in the 1987 Sotheby’s sale of the Duchess of Windsor’s jewels, only five were correctly attributed to her at the time. Belperron proves the rule from the other side — the premium attaches to documented authorship, and a stamp is simply the cheapest form of documentation.

What trades, within any of these houses, is narrower than the catalogue. The auction market is concentrated in pieces that are identifiable in a photograph from across a room, that carry a period the trade has a name for, and that have not been altered. A generic 1980s gold chain from a great house is worth its gold. An Art Deco bracelet with the house’s numbering and an archive letter is worth a multiple of its parts. The gap between those two outcomes has nothing to do with the quality of the gold.

No repeat-sale index exists for signed jewellery, and you should be suspicious of anybody who quotes you a return series for it. Art has Mei Moses and Artprice; wine has Liv-ex; watches have WatchCharts and Subdial, built from thousands of transactions in standardised references. Jewellery has none of that, for a reason that is structural rather than an oversight: almost every fine jewel is unique, so there is no reference to repeat. Two Art Deco Cartier bracelets are not the same asset in the way two Submariner 124060s are. The best the market can do is auction records for individual pieces, and auction records are the most survivorship-flattered data in the alternative-asset world, because the only pieces that reach a saleroom are the ones a specialist believed would sell.

What does exist is Knight Frank’s Luxury Investment Index, which tracks ten collectible categories including jewellery and coloured diamonds. The 2026 edition, published in April 2026, put the whole index at −0.4% for 2025 and +38.6% over ten years. For context on the same page, handbags returned −0.2% over twelve months and +85.5% over ten years, seventh of the ten categories. The index’s jewellery and coloured-diamond line items are not published in the material Knight Frank puts outside the report itself, we could not obtain them, and we do not quote them; treat any figure you see attributed to them without a date and an edition number as unverified. What Knight Frank does say in the 2026 commentary is that fancy coloured diamonds, which are under 0.1% of global diamond supply, held roughly steady through 2025, that blue stones were the standout, and that the year’s two largest fancy-coloured sales were the 9.51-carat Mellon Blue at $25.6 million and the 10.03-carat Mediterranean Blue at $21.5 million.

What the ten-year index number tells you plainly is that the whole luxury-collectible complex compounded at roughly 3.3% a year over the decade, against +313% for the S&P 500 on a trailing ten-year basis to August 31, 2026, or 15.2% a year. That comparison is before jewellery’s fees, insurance and the 28% collectibles tax, all of which sections ten to twelve price.

The record that does exist, and that is worth studying, is the record of exceptional single pieces — and the lesson in it is provenance rather than gemology. The Sotheby’s Geneva sale of the Duchess of Windsor’s jewels on April 2 and 3, 1987 made over $50 million against a pre-sale estimate of about $7.5 million. Elizabeth Taylor’s jewels made $137.2 million across 269 lots over two days at Christie’s New York on December 13 and 14, 2011 — the most valuable jewellery collection ever sold at auction, and part of a $156.75 million total for her whole estate — with her La Peregrina pearl necklace alone at $11.8 million against a $2 million to $3 million estimate. In each case the sale cleared several times the estimate, and the multiple was paid for a name attached to the owner on top of the name stamped in the metal. Provenance of that order is not an asset class. It is a lottery ticket that a few hundred objects on earth hold, and no strategy you can execute involves acquiring one.

One result from those collections is worth more than the rest of the list, and not for its size. The Duchess’s 1952 Cartier onyx-and-diamond panther bracelet made £4,521,250 — about $7.3 million — at Sotheby’s London on November 30, 2010, a record at the time for any Cartier jewel and for any bracelet at auction. The buyer was Cartier. The maison bid against the open market for a piece of its own 1952 production and took it at a record, which is the only public, priced proof anyone has of the argument in section one: the house buy-back desks are a real bidder with a real budget, not a museum courtesy.

Read it the way a bidder would rather than the way a press release does. Cartier Tradition is a commercial arm; it restores what it acquires and sells it on. So the underbidder in that room was not competing with an endowment, and the price was not charity — it was a trade buyer with the archive open in front of it paying what the archive said the piece was worth. Two things follow for you. On a documented piece from a defined period, there is a floor under the market that you cannot see on the screen and cannot easily outbid. And when a maison is the buyer of record, treat the result as the best comparable in the category rather than as a curiosity, because it is the one price in this market set by the party with complete information.

A laboratory-grown diamond is chemically, optically and physically a diamond; distinguishing it from a mined one takes equipment, not an eye. In 2016 a one-carat lab-grown stone traded at perhaps 20% below its mined equivalent. By 2026 the same stone trades at roughly $700 against roughly $4,200 for the mined equivalent — a discount of eighty to ninety per cent depending on grade. The wholesale series is the cleaner way to see it: a one-carat certified lab-grown stone went for about $3,400 at wholesale in 2020 and about $680 in 2026, a fall of four-fifths in six years, because the cost of growing one collapsed and the growers, concentrated in India and China, compete on nothing else. That is not a fashion. It is a manufactured perfect substitute arriving in a market whose entire pricing rested on scarcity, and no cartel can ration a factory.

The demand side moved as fast as the supply side. The Knot Worldwide’s 2026 Real Weddings Study, published on February 18, 2026 from 10,474 American couples married during 2025, put the lab-grown share of engagement-ring centre stones at 61% — past the 50% crossing of 2024, and a 239% rise since 2020. The average ring cost $4,600, down from $5,200 the year before, while the average centre stone grew to 1.9 carats — 2.0 carats at $4,300 for lab-grown, against 1.6 carats at $7,000 for natural. Buyers took a bigger stone for less money and did not care where the carbon came from.

Prices followed. Mined polished prices peaked in the first quarter of 2022 and have fallen since, with the damage concentrated in exactly the commercial sizes engagement rings use: small and mid-grade stones down by half or more, three-carat-plus and top-colour goods far less. IDEX’s own commentary puts its polished composite at almost 155 in March 2022 and at 90.02 in August 2025. Upstream, De Beers’ average realised rough price fell 7% to $142 a carat in 2025 and reached $105 a carat in the first half of 2026, 32% below the $155 of the first half of 2025, with the like-for-like rough price index down 16% (De Beers and Anglo American interim results, July 2026).

The company that built the market is being sold for something close to scrap value, and the dates matter because they mark the end of an eighty-year price-support regime. Anglo American, which took its De Beers stake to 85% in 2012, wrote the business down for the third time in three years on February 20, 2026 — a $2.3 billion pre-tax impairment that cut the carrying value to $2.3 billion from $4.1 billion, taking three years of writedowns to $6.8 billion — and has been trying to sell it since 2024.

By July 2026 Bloomberg reported advanced talks with the Global Diamond Consortium, led by the former De Beers chief Gareth Penny and including the governments of Namibia and Angola alongside large diamond traders, at about $1 billion for the 85% — reported as $750 million on completion and $250 million deferred, with a further sum said to be injected into the business. As of September 2026 Anglo’s chief executive Duncan Wanblad described the sale as in its final and most challenging phase, with completion targeted for the last quarter of 2026 and Botswana’s government among the approvals still to come. It had not closed. De Beers was valued at $17.6 billion when it was taken private in 2001.

What this did to jewellery, and the reason it belongs in a guide about signatures, is that it removed the floor under the unsigned piece. An unsigned diamond ring was always worth roughly its stones less a dealer’s margin; when the stones were appreciating, that was a tolerable place to be. Now the stones are a depreciating manufactured good and the ring has no other claim on value. A signed piece from a documented period still has the residual. The collapse did not raise the signature premium in dollars so much as it removed everything the signature was competing with.

Every gold jewel has a hard floor that a refiner will pay in cash on the day, and for the first time in a generation that floor is doing real work in the price. The arithmetic is public and takes ten seconds. A troy ounce is 31.1035 grams. Eighteen-karat gold is 75% fine by weight, fourteen-karat 58.3%. Divide the spot price by 31.1035 to get the value of a gram of fine gold, multiply by the karat fraction, multiply by the weight of the piece, and take off 5% to 10% for the refiner’s margin. That is the number below which nobody sane sells, and it is the number a dishonest buyer hopes you have not calculated.

Our tape closed gold futures at $4,443.90 an ounce on September 8, 2026 — spot was quoted around $4,411 the same morning — up 74.2% from $2,551.20 on September 12, 2024. At that price a gram of fine gold is $142.87, a gram of 18-karat is $107.16, and a plain 18-karat bracelet weighing thirty grams carries about $3,215 of metal before anybody discusses design. Two years earlier the same bracelet held $1,845. Nothing about the object changed. The composition of its floor did. Investing in Gold, on the precious-metals hub, argues the metal on its own terms; here it matters only as the number a refiner will pay you this afternoon.

Set the two halves of the jewellery case side by side and the inversion is complete. Over roughly the same window the metal roughly doubled and the commercial white diamond roughly halved. A piece that was 70% stones and 30% gold by value in 2021 may be the reverse in 2026, which changes what you should be willing to pay for it, what your insurer should be scheduling it at, and — this is the part owners miss — what a dishonest dealer can make by breaking it up. Estate jewellery is being scrapped for gold in volume — refiners reported a surge of unprocessed material during the spike that took our tape to $5,318 an ounce on January 29, 2026 before it fell back within a session — and every scrapped piece is one fewer example of a period that the market prices on scarcity. That is a slow, invisible supply squeeze on the signed vintage market, and it is the one genuinely bullish mechanism in this guide.

Our tape

Invest Alternative runs its own daily reads on this market and we present them as ours, not as market-wide returns. Our diamond series is the IDEX polished composite, which we read at 79.47 on September 8, 2026, from 79.38 on August 28 — nine observations, an early-history series and not a return index. Set against IDEX’s own published commentary, which puts the composite’s peak at almost 155 in March 2022 and its August 2025 close at 90.02, our reading sits roughly 49% below the high and about 12% below that August 2025 close. Our second gem series is the Fancy Color Research Foundation index, which we carry at −0.1% quarter on quarter in its April 2026 reading — one observation, a flat quarter at the top of the market while the ordinary trade fell. The FCRF’s own published first-quarter 2026 figure was −0.2% for the overall index, with pinks and blues each down 0.3% and yellows unchanged; the small difference is a reminder that ours is a single read of one vendor series and not the vendor’s own bulletin.

Diamonds and Gems is an “awaiting” category in our composite, carrying a 1.4% target weight and no live weight until the series has history; the IA Composite itself stood at 100.271 on September 8, 2026, up 5.74% over thirty days and 0.29% over a year, on a provisional basis. None of those is a jewellery-market number and we do not blend them with one.

Sapphires, rubies and emeralds of true gem quality are the only components of a jewel with a genuine long-run appreciation record, and that record is almost entirely a record of geography. A Kashmir sapphire, a Burmese (Mogok) ruby or a Colombian emerald with a laboratory report confirming origin and no treatment trades in a tier of its own; the visually similar stone from Sri Lanka, Mozambique or Zambia trades at a fraction. The market is paying for a nineteenth-century geological accident and for a Swiss laboratory’s willingness to certify it, and the reason the premium is durable is that the deposits are finished. Kashmir was mined for a few decades from the 1880s. There is no more of it. The supply is whatever survives in old jewellery, which is the purest form of the scarcity every other alternative asset only claims.

The record book is where the tier is visible. The Regent Kashmir, 35.09 carats, unheated and antique cushion-cut, made $9.5 million at Christie’s Hong Kong in May 2025 — $271,515 a carat, a world record price per carat for a sapphire, beating the $243,703 a carat set by the 27.68-carat Jewel of Kashmir at Sotheby’s Hong Kong in October 2015. In rubies, the Sunrise Ruby, a 25.59-carat Burmese stone in a Cartier ring, made $30.3 million at Sotheby’s Geneva on May 12, 2015 and held the coloured-stone record until the Estrela de Fura, 55.22 carats cut from a 101-carat rough, made $34.8 million at Sotheby’s New York on June 8, 2023 — the record for any coloured gemstone at auction. The detail an investor should notice is that the Estrela de Fura is Mozambican. The Mogok premium is real, and the market will still pay for size and colour from anywhere when the stone is genuinely extraordinary.

Fancy coloured diamonds sit above even that, and they are the one part of the diamond business a laboratory has not undone. The CTF Pink Star, 59.60 carats and internally flawless fancy vivid pink, made $71.2 million at Sotheby’s Hong Kong on April 4, 2017 and remains the most expensive gem ever auctioned; the Williamson Pink Star, 11.15 carats, made $57.7 million in Hong Kong on October 7, 2022 at a record per carat and stands second; the Oppenheimer Blue, 14.62 carats, made CHF 56.8 million, about $57.5 million, at Christie’s Geneva in May 2016; the Mediterranean Blue, 10.03 carats and cut from a 31.94-carat Cullinan rough, made $21.5 million at Sotheby’s Geneva on May 13, 2025. The supply argument is unusually clean. The Argyle mine in Western Australia, source of roughly 90% of the world’s pink diamonds, closed in November 2020 after thirty-seven years and more than 865 million carats of rough, and no comparable deposit exists. Laboratories can grow pink and blue diamonds; they cannot yet grow them in the size, saturation and purity the auction tier requires, and the trade keeps the two markets apart with origin reports.

Read that chart carefully before you take the wrong lesson from it. Every one of those stones is a museum object bought by one of perhaps fifty buyers on earth, and none of them tells you what a good five-carat Burmese ruby does over a decade. What the chart establishes is the direction of the premium — origin, saturation, size, no treatment — and the fact that the tier kept setting records through the same four years in which the commercial diamond trade fell apart. For a buyer without eight figures, the executable version is one stone of unquestioned origin with top laboratory reports and no treatment, bought at auction with the premium priced in and held for a decade. It is a connoisseur’s market and it punishes the amateur exactly as art does.

A grading report is a description, not a valuation, and the difference is where most amateur money is lost. The Gemological Institute of America, founded in Los Angeles in 1931 by Robert Shipley, coined the four Cs in the early 1940s and turned them into the International Diamond Grading System in 1953 — carat, colour, clarity, cut, the four attributes the diamond trade prices on — and a GIA report is the document a dealer will want for any white diamond of consequence. What it does not do is tell you what the stone is worth, because each of those attributes sits on a continuous price curve and the curves interact. Two stones with identical certificates can differ 20% in value on cut quality, fluorescence and how the inclusions actually read to the eye. A certificate is a floor on the description and nothing else.

For coloured stones the report matters more, because origin and treatment are the price. Three Swiss and American laboratories carry the trade’s trust on that question: SSEF in Basel, Gübelin in Lucerne and the American Gemological Laboratories in New York, with GRS also widely quoted. Their reports address two things a diamond report does not. The first is geographic origin, determined from trace-element chemistry and inclusion suites and stated as an opinion rather than a fact. The second is treatment: whether a sapphire has been heated, whether a ruby has been lead-glass filled, whether an emerald’s fractures have been filled with oil or resin and at what degree — minor, moderate, significant. The gap between an unheated Burmese ruby and a heated one of identical appearance is not a discount, it is a different market.

Two practical rules follow, and neither is expensive. Buy the report before the stone if you can — reputable sellers of significant stones already have one, and a seller who resists an independent report on a stone above a few thousand dollars has told you something.

And expect to pay for it, because the tariffs are published and they scale with carat weight and with the questions asked. SSEF’s price list runs from CHF 330 for an identification report on a stone under two carats to CHF 4,200 for identification and origin on a stone of 100 to 150 carats, with a 50% surcharge for three-day turnaround and CHF 650 a year for its preferred-client rate. Gübelin quotes roughly CHF 250 to CHF 2,500 for a full gemmological report and CHF 200 for its cut-down Gem Passport on an unmounted emerald, ruby or blue sapphire up to three carats. GIA’s coloured-stone schedule is cheaper still — on the order of $115 for identification on a 30-to-50-carat stone, with about $100 more for origin in the same band — and from January 2026 it extended origin determination to opal, peridot and demantoid garnet. AGL is the exception: it releases its price list on request rather than publishing it, so budget from the Swiss schedules and ask before you send. Those are 2026 figures against a small stone and rounding against a large one, and every laboratory revises them, so confirm the current tariff before you set a bid.

Copying a jewel convincingly means buying the actual stones, and at that point the counterfeiter has spent most of the money. That single economic fact is why jewellery has no superfake problem and why authentication here works better than it does in handbags or watches. The frauds in this market are subtler: a genuine period piece with replaced stones, two damaged pieces married into one, a modern piece with a forged stamp, or a real piece whose date has been quietly moved forward thirty years in the description.

Three layers of evidence answer those questions. The first is the hallmark — the assay marks struck into the metal. In France an eagle’s head has marked 18-karat gold since 1838 and a lozenge-shaped punch carries the maker’s mark. In the United Kingdom hallmarking is compulsory on any gold article over one gram, and the assay offices at London, Birmingham, Sheffield and Edinburgh each strike an identifiable town mark — leopard’s head, anchor, rose, three-turreted castle — alongside a date letter, so a British hallmark on a vintage piece gives you a year; the date letter has been optional rather than compulsory since 1998, which matters only for modern work. In the United States there is no compulsory hallmarking at all: the National Gold and Silver Stamping Act of 1906 requires that a quality mark, if applied, be accurate within a fixed tolerance, and since a 1961 amendment that it sit beside a registered trademark, which means a bare American piece tells you nothing and a marked one is only as good as the maker’s honesty. A specialist reads all of this in seconds and you can learn the main ones in an afternoon.

The second layer is the house numbering. Cartier, Van Cleef & Arpels and Bulgari have numbered their production for most of a century, and the number is the key that opens the third layer: the archive. Cartier has run a vintage arm, Cartier Tradition, since 1996, which researches pieces against the archive, restores them and issues a report on materials, design history and every step of the restoration — though it does not disclose previous owners. Van Cleef & Arpels runs a Heritage collection begun in 2007, whose buyers work with the Maison’s Patrimony department through client files, account books and sketch libraries to authenticate a piece and issue its certificate. An archive letter that ties a specific number to a specific year and a specific original owner is the single most valuable piece of paper in this market — it is what converts “an Art Deco bracelet in the Cartier manner” into “a Cartier bracelet, 1928, sold to…”, and the price difference between those two catalogue entries is routinely a multiple rather than a margin. The houses’ archive services are slow, discretionary and free or near-free, and they are under no obligation to help you.

The practical consequence is that you should treat documentation as part of the asset and buy it with the piece. A signed piece with its original fitted case, the house’s numbering intact and an archive letter is a different security from the same piece without them, in the same way that box and papers change a vintage watch. Buy the paperwork or budget the months it takes to chase it.

Price in this market is discovered in four venues that barely talk to each other, and knowing which one you are standing in is most of the skill. The auction houses — Christie’s, Sotheby’s, Phillips, Bonhams and a long tail of regional salerooms — are the only place a price becomes public, which is why every comparable a dealer quotes you traces back to a saleroom. They hold jewels sales in New York, Geneva, London and Hong Kong on a seasonal calendar, they publish estimates the house itself chose, and they sell against a confidential reserve. If you have read How Art Auctions Work you already know the machinery, and it is identical here: the estimate is a marketing figure, the reserve is confidential and, in Sotheby’s own words, is generally set at a percentage of the low estimate and will not exceed it, and a lot that fails to sell becomes public knowledge that follows the piece for years.

The dealers and estate jewellers are the second venue and the largest by volume. They buy from families, from the trade and at auction, and they sell at a marked-up retail price to a walk-in client. A dealer’s bid is where a private seller actually discovers what a piece is worth in cash today, and it is a long way below the auction result the seller read about, because the dealer is buying inventory that may sit for two years and is pricing that risk. Published trade guidance puts a dealer’s cash bid on a saleable signed piece somewhere between 30% and 60% of what the same dealer expects to ask — and for ordinary estate goods with no name on them, at only 10% to 25% of what the piece originally cost at retail, which in practice is a scrap bid with a little sentiment attached.

The maisons themselves are the third venue, and the one most outsiders mis-file. Cartier Tradition, running since 1996, and Van Cleef & Arpels’ Heritage collection, begun in 2007 and worked by the Patrimony department, both buy their own historic production back — typically from clients, sometimes from dealers, and by their own account rarely from auction houses, though the 2010 panther bracelet in section three is a public and expensive exception. What they do with it is the point: they research it against the archive, restore it, and then exhibit it or sell it again. Van Cleef showed some forty Heritage pieces at TEFAF in March 2026 out of a standing selection of about 150 held for sale.

That makes the house a commercial buyer rather than a museum, and it changes how you should treat it at both ends of a trade. As a seller you are dealing with the best-informed bidder in the market, one that pays well, pays quickly and does not haggle much, because it is stocking a retail operation with pieces it can authenticate from its own archive in a way no dealer or saleroom can match. It is also selective to the point of eccentricity and will not tell you its criteria. If your piece is important, a call to the heritage department before you consign is free and occasionally worth a great deal. As a buyer you are on the other side of that: on the best-documented pieces from a defined period, the maison is the competing bid, and the price it sets is the one the rest of the market will quote back at you for a decade.

Volume across the salerooms is real but poorly disclosed. Sotheby’s reported $2.7 billion of luxury sales across all its luxury departments in 2025, up 22% (The Art Newspaper, December 17, 2025); Christie’s luxury business in the same year is reported two ways that we can now reconcile: The Art Newspaper’s year-end tally on December 17, 2025 put it at $795 million, up 17%, while Christie’s own subsequent full-year statement put the luxury division above $1 billion excluding private sales, on a 90% sell-through and hammer prices at 129% of low estimates. Neither house publishes a jewellery-only total, which is why none appears anywhere in this guide.

The fourth venue is online, and it has grown fastest. The April 2020 Tutti Frutti sale in the cold open was an early proof that a seven-figure jewel could sell to somebody who had never held it, and the timed online sale is now where the houses clear most of the lower and middle market. On the general marketplaces, eBay’s Authenticity Guarantee is mandatory and free on jewellery sold at $500 or more, optional at $40 between $200 and $499, and available only on deliveries inside the United States — a real improvement on the position ten years ago, and still not a substitute for a laboratory report.

Behind all four sits the same demand engine, and it is worth knowing that the business is healthy even where the asset is not. Richemont’s Jewellery Maisons — Cartier and Van Cleef & Arpels — grew sales 14% at constant rates to €16.5 billion in the year to March 2026 on a 30.5% operating margin, after 11% the year before, in a period when the group’s watch division shrank and Bain’s industry study singled out jewellery as the category that held up best as leather goods fell. Strong primary sales are good news for the signature’s cultural standing and bad news for anybody hoping current production will appreciate: the houses are making more of it, not less.

Transaction friction is the largest single determinant of return in this asset class, larger than any view you might have on gold or on Art Deco, and it is charged on both sides of the trade. At auction the buyer’s premium is added to the hammer price and paid by you. Both major houses now charge 28% on the first tranche of a lot: Sotheby’s from February 13, 2026, on the first $2 million in New York, and Christie’s from September 1, 2026, having moved to 27% up to $1.5 million a year earlier. Both then charge 22% from $2 million to $8 million and 15% above that, so almost nothing an ordinary buyer bids on reaches the second tier. Phillips charges 29% on the first $1 million, with a 25% rate available to a bidder who lodges a binding written bid at or above the low estimate 48 hours before the sale — a structure it introduced on September 1, 2025 for every category except watches. Bonhams went further on October 1, 2026, to 30% on the first £25,000 or $35,000 — the highest low-tier rate any of the four has published, and its first change since the 2023 schedule of 28% on the first £40,000 that it replaced. Its new ladder runs 28% to £500,000, 25% to £1 million, 21% to £5 million and 14% above, and it excludes wine and whisky, coins and medals, and motor cars.

Four increases across the houses in the thirteen months from September 2025 to October 2026 — Christie’s twice, Sotheby’s once, Bonhams once, with Phillips restructuring alongside them — all in the same direction, is the fact to carry forward from that chart rather than any individual rate. Assume the premium you pay in five years will be higher than the one you pay today, and check the schedule in the back of the catalogue every single time, because it changes between seasons and it differs by saleroom.

On top of the premium sits sales tax, charged on the premium-inclusive price at the rate where the lot is delivered — 8.875% in New York City, typically 6% to 10% across the states that levy it. That money is basis, not loss, but it is cash out the door and it raises your break-even by most of a tenth before you own anything.

When you sell, the seller’s commission is nominally around 10% of the hammer, and the published schedule tells you plainly where the house’s attention sits. Sotheby’s waives the commission outright on any consignment with a low estimate of $5 million or more. Between $20 million and $50 million the arrangement inverts: the seller takes 40% of the buyer’s premium on top of the hammer, so the house is paying the consignor for the privilege of selling the piece. Nothing in this guide is aimed at a consignment of that size, and that is the useful part. If the rate can go to zero and then below it at the top, the 10% quoted to you at the bottom is a starting position rather than a price. Expect a menu of smaller items alongside it — insurance while the house holds the piece, photography, catalogue illustration, and a fee if you withdraw the lot after consigning. Ask for all of them in writing before you sign, and get a second house’s terms sheet before you answer the first.

The alternative to consigning is selling to the trade, and the trade-off is speed against price. A dealer pays cash this week at 30% to 60% of what they expect to ask; an auction pays in six weeks to six months at a hammer that may be higher or may be nothing at all, and a failure is public. For a piece worth less than about $10,000 the auction route usually does not survive the fees; for a piece with an archive letter and a period the market wants, it usually does.

Jewellery is the cheapest of the physical alternative assets to hold, which is a genuine advantage over cars, wine and art, and one this category rarely gets credit for. Cheap is not free, and the three line items are insurance, appraisal and servicing.

Insurance is the big one and the one most owners get wrong. A standard homeowner’s or renter’s policy caps jewellery at a low sub-limit, typically $1,500 to $2,500 for the whole category, and pays actual cash value rather than replacement. Anything you regard as an asset has to be scheduled: listed individually on a personal-articles floater with an appraisal, insured at agreed value, and covered for theft, loss and mysterious disappearance worldwide. That runs roughly 1% to 2% of insured value a year — consumer guides put the whole published band at 0.5% to 3%, and the high-net-worth carriers sit at the bottom of it if the rest of your business is with them. On a $30,000 piece, $300 to $600 a year. On a $100,000 stone, $1,000 to $2,000.

Appraisals have to be refreshed for the insurer, usually every two to three years, at roughly $100 to $300 each, and in a market where gold has moved 74% in two years an out-of-date appraisal is a real exposure: you are paying premium on a number that no longer replaces the piece. Servicing is cheap and consequential. Signed jewellery should be serviced only by the house that made it, because third-party repair voids both the warranty and, in practice, the archive’s willingness to stand behind the piece; a re-tipped claw done by a good bench costs very little and a stone lost from a worn setting costs everything.

Storage asks almost nothing: a safe, away from light and household chemicals, with pieces separated so harder stones do not abrade softer ones. Gold and platinum do not corrode; pearls and opals want humidity and will craze in a bank vault; emeralds are fracture-filled and must never see an ultrasonic cleaner. Call the all-in carry 1.5% to 2% a year on a scheduled piece, against roughly 5% for a classic car and 2% to 3% for a cellar. That is the strongest practical argument for this category over its physical rivals. It is also why the return in jewellery is decided by the fees and the tax rather than by the carry, which is what the next two sections cost out.

A jewel held as an investment is tangible personal property, and the United States taxes its gain as a collectible. Held more than a year, the gain is taxed at a maximum federal rate of 28% under IRC §1(h)(4), against 20% for securities, and the 3.8% net investment income tax under §1411 applies on top for taxpayers above $200,000 of modified adjusted gross income filing single or $250,000 filing jointly. The effective top federal rate is therefore 31.8%, before any state tax. Held a year or less, the gain is ordinary income at your marginal rate. There is no long-term rate below 28% for a collectible and no way to elect into one.

The asymmetry is what costs people money. If the piece was personal-use property — you wore it — a loss on sale is not deductible at all under §165(c). If it was investment property, bought and stored and never worn, the loss is a capital loss usable against gains and against up to $3,000 of ordinary income a year. The IRS decides which by conduct, and the conduct that makes a jewel enjoyable is the conduct that makes its loss disappear. You are taxed on the way up and disallowed on the way down, and “I can always wear it” has a specific price.

Three more rules close the obvious escapes. The wash-sale rule in §1091 applies to stock and securities, not collectibles, so you can harvest a loss on an investment-held piece and rebuy immediately — a rare point in the buyer’s favour. Like-kind exchanges under §1031 have been limited to real property since 2018, so you cannot roll a Cartier gain into a Van Cleef. And the One Big Beautiful Bill Act of July 4, 2025, restored the Form 1099-K reporting threshold to $20,000 and 200 transactions, retroactively per the IRS’s own FAQs, which means a casual seller below that line receives no form and still owes the tax.

Two details are worth planning around. Basis includes everything you paid to acquire and authenticate: hammer price, buyer’s premium, sales tax, shipping, insurance in transit and laboratory fees. Keep the invoices; the paperwork that lifts the resale price is the same paperwork that proves the basis. And jewellery held at death receives a stepped-up basis to fair market value in the estate, which is the only clean way this asset class ever escapes the 28%. A piece bought at $30,000 and worth $60,000 at death passes to an heir with a $60,000 basis and no tax on the gain, which is why so much serious jewellery is never sold by the person who bought it. If your holding period is genuinely indefinite, §1014 makes never selling the tax-efficient plan rather than the lazy one.

Take a signed vintage piece bought at auction in New York, held ten years, insured properly, and consigned back to the same house. Every figure below is arithmetic on the fee schedules already cited, and the point of the exercise is not the answer but the size of the hurdle.

You bid to a $23,500 hammer. The buyer’s premium at 28% adds $6,580, so the piece costs $30,080 before tax. New York City sales tax at 8.875% on the premium-inclusive price adds $2,669.60, and shipping with transit insurance adds $150. Your basis is $32,900.

Now hold it. Insurance at 1.5% of a scheduled value starting at $30,080 and revalued upward 3% a year costs $5,173 over the decade. Three appraisal refreshes at $250 each add $750. Total carry: $5,923. Your total cash out of pocket after ten years is $38,822, and you have owned a beautiful object, which is worth something the spreadsheet cannot hold.

Now sell. You consign to auction, negotiate the seller’s commission to 10% of hammer, and pay federal tax at 28% plus the 3.8% net investment income tax on the gain over basis. Solve for the hammer price at which you get your money back and the answer is $46,204. At that hammer the house adds its 28% premium and the next buyer pays $59,141. Hold the two multiples apart, because they are the whole lesson: that is 1.97 times the $30,080 the piece cost you with premium, and 1.80 times the $32,900 basis you actually carried. The next buyer has to find nearly double what you paid at the rostrum for you to come out level, and the gap between 1.97 and 1.80 is the ten years of insurance, appraisal and tax sitting in between. Expressed as a return on the hammer you paid, break-even requires 7.0% a year for a decade. Add New York State and City income tax at roughly 5% and the break-even hammer rises to about $46,970.

Three readings of that chart, in order of importance. First, the friction is the investment case. Nothing in the example assumes a bad market; the piece could rise 97% over a decade and hand you exactly nothing. Second, the same money in equities was not close. The S&P 500 returned 313% over the ten years to August 31, 2026, or 15.2% a year, with daily liquidity, no consignment fee, no insurance and a 20% top federal rate on the gain instead of 28%. Third, the exits are not equivalent. Sell to a dealer instead of consigning and you skip the wait and the risk of a failed lot, but a bid at 30% to 60% of the dealer’s ask will not reach $46,204 on a piece the dealer expects to ask $55,000 for. The auction is the expensive exit that pays more, and the dealer is the cheap exit that pays less; there is no third door.

Five failures account for almost all of the money lost in this market, and none of them is a price forecast. The first and largest is the altered piece. A genuine Art Deco Cartier bracelet whose original diamonds were swapped for modern stones of the same size is still a Cartier bracelet and is worth a fraction of an untouched one, because the trade prices originality and a specialist can see a modern brilliant in a 1928 setting from a metre away. Its cousin is the married piece: two damaged jewels combined into one saleable object, often decades ago, often by a competent jeweller. Neither is detectable from a photograph. Both are detectable by a specialist with a loupe in ninety seconds, which is the entire argument for never buying an important piece unseen and unadvised.

The second is treatment misrepresentation on coloured stones. Heat, lead-glass filling, diffusion, oiling and resin all change what a stone is worth by multiples, all are invisible without equipment, and all are entirely legal to sell as long as they are disclosed. The market for undisclosed treatment is not a fraud ring; it is a chain of resellers each of whom accepted the last one’s word. A current report from SSEF, Gübelin or AGL in your own name is the only defence, and it is cheap relative to what it protects.

The third is the failed lot. Consign at a reserve set to protect a price you have in your head, fail to sell, and you have established a public negative comparable that follows the piece on every price database for years, at no charge that appears on any terms sheet. The trade convention is that a bought-in lot should rest for several seasons before it is offered again, which means a single greedy reserve can freeze your capital for half a decade.

The fourth is title and theft. Jewellery is the most stealable asset in this series and one of the hardest to trace, and stolen goods carry no good title however innocent the buyer. Check the Art Loss Register or an equivalent database on anything significant — a single search costs about $110 plus VAT against a private database of some 700,000 lost, stolen and looted objects, and the certificate’s real value is as evidence of due diligence rather than proof of clean title — keep photographs and a written description with the serial or inventory number, and store both away from the piece. An insurer will ask for exactly this after a loss, and the owner who cannot produce it is arguing about the settlement rather than receiving it.

The fifth is liquidity, which is worse than it looks. There is no bid on a jewel the way there is a bid on a gold coin. Between the dealer’s cash offer and the auction’s uncertain hammer sits a gap of tens of per cent, and in a soft season the auction end of that gap can vanish for a year. Anybody who tells you jewellery is liquid is describing gold content, which is liquid, and mislabelling the part you actually paid for.

The sequence below is the one we would follow with our own money, and its first three steps cost nothing.

1. Pick one house and one period and read a year of results. Take the Artnet Price Database, Artprice or the houses’ own archives, choose Cartier in the 1920s and 1930s, or Van Cleef mystery-set pieces, or Bulgari of the 1960s, and read every jewels-sale result for twelve months: estimates, hammers, bought-ins, condition notes. You will learn the clearing price for the pieces you would actually want, which is a different number from the records this guide quotes.
2. Go and handle a hundred pieces. Attend the pre-sale views in New York, Geneva or London — they are free, open to anyone and staffed by specialists who will answer questions. Handle real examples and fakes side by side until period workmanship stops being an abstraction.
3. Learn the four hallmark systems. French, British, Italian and American conventions cover most of what you will see, and knowing them turns an unattributed piece in a provincial saleroom into a dated one.
4. Set your budget in hammer terms, not in total terms. Decide what you will pay all-in, then divide by 1.39 — the 28% premium and 8.875% New York sales tax compounded — to get your bidding ceiling. Write the ceiling down before the sale and do not move it in the room.
5. Buy your first piece for between $5,000 and $15,000, at auction, with condition report in hand. Request the house’s written condition report on every lot you are serious about, and ask specifically whether stones appear original to the setting. Their answer is a document you can rely on later.
6. Chase the archive. The moment the piece is yours, apply to the maison’s heritage department with the number and the photographs. It may take months, it may fail, and if it succeeds it has materially changed what the piece is worth.
7. Schedule it and photograph it. A personal-articles floater at agreed value, a written appraisal, photographs from six angles, and the file kept somewhere other than the house holding the jewel.
8. Decide your exit before you need one. Two written offers when you sell — one auction estimate with the terms sheet, one dealer bid in cash — and a reserve set at your walk-away number rather than the house’s low estimate.

Five readings would change our view, and each has a threshold rather than a direction.

The polished-diamond index

Our tape read the IDEX composite at 79.47 on September 8, 2026, roughly 49% below its March 2022 peak of almost 155. A sustained move back above 100 would mean the natural stone has found a floor above the lab-grown cost curve and that unsigned jewellery has value again; a break below 70 would mean the commercial trade is still being repriced and that the signature is doing even more of the work than this guide assumes.

The lab-grown share of engagement rings

61% of 2025 purchases (The Knot Worldwide, 2026 Real Weddings Study, February 2026), up from a crossing of 50% in 2024. A reading above 75% in a future edition would confirm the natural stone has become a niche luxury rather than a default; a reversal below 50% would be the first evidence in a decade that consumers price origin.

The De Beers sale

Anglo American targeted completion in the last quarter of 2026 at about $1 billion for its 85% stake, subject to Botswana’s approval, against a $17.6 billion valuation when the company was taken private in 2001. Watch the closing price and the new owner’s marketing spend: a buyer who restarts category advertising at scale is defending the natural stone’s story, and a buyer who does not has bought a mining asset and let the story go.

Gold

Our tape closed gold futures at $4,443.90 an ounce on September 8, 2026, having spiked to $5,318 on January 29, 2026 and given it all back in a single session. The metal floor under every gold jewel moves with that number, and so does the rate at which estate pieces are scrapped. Gold sustained above $5,000 — which January was not — would accelerate the destruction of ordinary period jewellery and tighten the supply of the signed pieces that survive; gold back below $3,000 would strip roughly a third out of the metal content of every 18-karat piece and take the support out from under the bottom of this market.

The buyer’s premium

Both major houses reached 28% on the first tranche during 2026, and Bonhams went to 30% on October 1 — four increases across the trade in thirteen months, and the first low-tier rate above 29%. The threshold to watch is either major house following Bonhams to 30% on the first tranche. That would push another slice of the middle market — the $5,000 to $25,000 pieces where the fees already barely work — out of the salerooms and into private treaty and dealer inventory, where prices are not published and the outsider’s information disadvantage is largest.