The Record is built in the open. We have read the auction record since the 1970s, and it all starts from one line — the Central 80%, the figure of record. No black box. Here is the method.
The dataset
A single canonical record, since 1975.
The index is grounded in a single canonical dataset of public auction sales of fine art and related categories.
| Coverage | 1975 (or later) to the latest observed month |
|---|---|
| Scope | Public auction sales reported by the recorded houses across the principal international markets — UK, US, France, Germany, Switzerland, Italy, Hong Kong, China, Japan and others |
| Mediums | Unique two-dimensional works, unique three-dimensional works, editions (prints, multiples), photographs |
| Currency | All values normalised to GBP at the prevailing rate at the date of sale; original currency retained for reference |
| Prices | All buyer’s premiums are stripped back to the hammer price — the figure the lot cleared for — so every sale reads on the same basis |
| Exclusions | Withdrawn and unverifiable sales are excluded. Private treaty, gallery, and dealer sales are out of scope — a transaction that cannot be independently observed, dated and reproduced cannot be used as evidence |
Each assessment records the dataset’s latest observed month and is reproducible against that snapshot. The dataset is updated on a regular cycle, and historic assessments remain reproducible against the data and methodology in force at the time of the case.
The index of record
The Central 80% Index.
This is the price record for a single artist — the typical value of their work at auction, tracked over time. Sales are stripped of buyer’s commission and converted to one currency (GBP) on the day of sale (or nearest date thereafter). For each month it pools the trailing twelve months of sales, discards the lowest and highest tenth, takes the mean of the central 80%*, applies a short recursive smoother, and rebases.
The line is, in effect, the world’s salerooms reported on a single scale. It reads the central tendency of what actually cleared — every qualifying public auction sale in a medium category, with no eligibility filter, no synthetic work, and no imported data that cannot be checked. It is the line of record: a UK tax authority has relied on it in chattels casework for over two decades, and its historic values are never restated — a value read in 1982 still reads the same today.
* The Central 80% methodology was built with the London School of Economics.
The evidence
Every figure traces to a real lot.
The lines and reads are built from named auction results, not estimates. When the trend is gated, the instrument hands you the relevant recent comparables; you can drill into the recent lots — the house, the sale, the price hammered.
Individual artists and groups
An index for each artist the record can carry.
An artist is given their own Central 80% index once their auction record holds enough qualifying sales to read a stable central tendency. Below that threshold the line would track sampling noise rather than the market, so no index is published — the record only speaks where it has the evidence to.
Each individual artist index is also grouped into indices by period or school, so a single name can be read on its own or against the movement of its cohort. Where an artist or group is not yet covered and the record can support one, we accept requests to add them.
Compare
Markets, side by side.
Set whole markets side by side. Rebased to a common base, the record measures how an artist’s — or a group’s — market moved relative to others across decades.
Lay a group index against individual artists, or set up to three markets side by side — each indexed to 1000 at a common date, so the chart reads as pure relative performance: who pulled ahead, who fell behind. Comparisons are held to the same medium (unique works against unique works, never against prints). Rebased this way, the record becomes raw material for academic and institutional research — tracking how a whole school moved, and how its names diverged, across decades.
What it does not address
Disclosed on every assessment.
Each of the limitations below is disclosed on every assessment under Important Notes.
- Auction-realisable, not retail or gallery. Retail and replacement values are typically higher. Practitioners using an assessment for a non-auction question must apply the appropriate adjustment.
- No model of condition, provenance, quality, or attribution. The methodology operates at the artist–medium level. A work in unusual condition, with unusual provenance, or whose attribution is contested may clear at a value that diverges materially from the indicative range.
- No model of sale context. Lot positioning, marketing investment, third-party guarantees, and the level of competing material in the sale are not modelled.
- Coverage boundaries. Works outside the principal medium categories, sold through routes outside the captured auction houses, or by artists with no auction record, are out of scope.
- Not a forecast. The methodology describes observed movement between two real dates. It is not a prediction.
- Indicative range, not a guarantee. The reported range expresses variance, not guaranteed clearance.
- Authenticity is assumed. A work whose authenticity is in question is outside the scope of any market-evidence valuation.
Auction House Sample
A sample that moves with the market.
The record is built from public auction sales at a defined, evolving set of houses. From 1975 to October 2007 the data was drawn from the Art Sales Index (ASI), whose coverage leaned toward European regional salerooms; AMR has collected the record itself since.
By then the auction market had already become far more international, so the emphasis moved away from the smaller regional salerooms and toward the jurisdictions where the market had grown — South Korea, Singapore and India, alongside Australia, South Africa and Mexico, and others.
The sample of houses is kept under continual review, with the emphasis on collecting where established and recognised emerging artists sell. New artist names are added to the database regularly.
Versioning
Each saved case stamps the methodology version.
Every saved case is stamped with the dataset snapshot and method revision it ran under, and stays reproducible against the version then in force. The published Central 80% Index itself is never revised — a value read in 1982 still reads the same today; only the dataset extends forward each cycle. The durable-trend read layered on top (the AMR Trendline) is versioned separately and documented on its own page. Material changes to the dataset, the rule, or the output structure ship as new revisions with a changelog.
The method, in the open
Documented in full — and citable.
This page documents The Record— the Central 80% — end to end: the dataset, the central-80% method, the auction-house sample, and the limitations disclosed on every assessment. It is open by design, so any figure read from it can be cited and checked. The durable-trend read layered on top — the AMR Trendline— is documented on its own page; its full methodology paper and equation-level technical companion are available on request.
The AMR Central 80% Index — the figure of record, published continuously since the 1970s · the central-80% method built with the London School of Economics.