Art Market Research: Which Data Sources to Trust

⁰¹Art market research draws on a handful of source types, and the most common mistake is treating them as interchangeable. A transaction record, an estimate built from a survey and a journalist's summary of a sale are different kinds of evidence, even when they report the same number.

⁰²A useful first sort is to ask whether a source is primary (a record of an actual transaction), derived (an aggregate or index built from records), estimated (an extrapolation from a sample) or interpretive (commentary on any of the above). Each layer adds usefulness, and each adds assumptions.

⁰¹Price databases that compile public auction results are the backbone of most artist-level research. They are searchable, cover many houses and go back years, and every entry refers to a real, dated sale.

⁰²Their limits are structural. They record only public auction sales, so the primary market and private sales are absent. Coverage varies by house and region, with smaller and regional salerooms less consistently captured. Conventions differ on whether prices include the buyer's premium and on how bought-in lots are recorded, and guarantee arrangements are generally not captured at all.

⁰³Entry errors also occur. Dimensions, media, attributions and dates can be transcribed wrongly, and the same work can appear more than once. For any conclusion that matters, the underlying catalogue entry is worth checking.

⁰⁴Used well, a database is a starting point for a list of comparables rather than the answer. Filtering carefully by medium, size, date and subject, reading each catalogue entry, and noting any guarantee symbols or condition issues turns a list of prices into evidence. An average across every result for an artist, by contrast, mixes works that have little in common.

⁰¹Industry reports, often published annually with the support of an art fair, bank or professional services firm, are the main source of market-wide estimates. They typically combine auction data with surveys of dealers, collectors or wealth managers, and they are frequently the origin of the headline figures quoted in the press.

⁰²Surveys depend on who responds. A dealer survey reflects dealers willing to share their data; a survey of wealthy collectors reflects those the sponsor can reach and who choose to answer. Response rates and sample sizes are not always prominent, and methods can change between editions, which weakens year-on-year comparisons.

⁰³Sponsorship does not make a report wrong, but it is part of the context. A report backed by an institution with an interest in art as an asset class, or in fair attendance, is worth reading alongside one with different backers. Wealth reports that track art among other luxury assets, for example, sit within a broader investment narrative.

⁰⁴Within a single report, it helps to separate figures that were measured from figures that were modelled. Auction totals compiled from public results are closer to measurement; estimates of dealer sales, private sales or collector behaviour extrapolated from survey responses are modelled. Reports generally explain which is which in a methodology section, and that section is often more informative than the summary.

⁰¹Indices attempt to turn scattered sales into a single trend. Repeat-sales methods follow works that sold more than once; hedonic methods model prices from characteristics such as size, medium and date. Both rely on auction data and inherit its gaps.

⁰²Repeat-sales indices only see works that came back to market, which may not represent the ones that did not. Hedonic models depend on which characteristics are included and can miss quality differences a specialist would see at once. The site's piece on artist price indices covers these issues in detail; for research purposes, the key is to know which method sits behind any index being cited.

⁰¹Most people meet art market data through the press, and much press coverage originates in announcements from auction houses, fairs and report publishers. Those announcements are accurate about what they state and selective about what they emphasise: records, sell-through rates and totals appear; unsold lots, guarantees and private-sale definitions appear less often.

⁰²Precise wording matters. An auction record for an artist is not an all-time record that includes private sales; a sale total may be premium-inclusive while the estimates quoted beside it are not; a lot 'sold above estimate' may be a premium-inclusive price compared with a hammer-based range. None of these is misleading when read carefully, but each is easy to repeat carelessly.

⁰³That is not a reason to ignore coverage. It is a reason to trace any figure back to its original source before relying on it, and to notice when a secondary report has dropped a qualification the original included.

⁰¹Before any piece of art market analysis draws a conclusion, it should be able to answer a short set of questions about its sources.

⁰²No source passes every test, and that is acceptable. The discipline is to state the limits alongside the finding. A conclusion framed as what auction results for an artist show, excluding private sales, sounds weaker than one that simply says what the market shows. It is also far more reliable.

⁰³Where two credible sources disagree, the disagreement is itself informative. It usually points to a difference in coverage or method that is worth understanding before choosing between them.