When an appraiser tells you a painting is worth $400,000, that number almost always comes from one method: comparable sales. The appraiser assembles a small set of recently sold works that resemble the one in front of them, adjusts each sale price up or down for the ways those works differ, and reconciles the adjusted figures into a single fair market value. It is the same logic a home appraiser uses on a house, applied to an asset where no two units are ever identical. The professional standards bodies treat it as the default. A 2024 methodology guide aligned with International Society of Appraisers practice calls the sales comparison approach the strongest indicator of value whenever an active secondary market exists. The whole exercise rests on one economic idea, the principle of substitution: a buyer will not pay more for a work than the cost of acquiring a reasonably comparable one.

We spend a lot of time on exactly this question, because before we buy a painting we have to fair-value it, and the gap between that fair value and the asking price is the entire edge. So it is worth walking through how the estimate gets built, where the adjustments come from, and where the method quietly breaks down.

How do appraisers select comparable sales for a work of art?

The first move is to define the value problem and the relevant market, then go looking for arms-length sales of works that are as close as possible to the subject. Under the Uniform Standards of Professional Appraisal Practice (USPAP), an appraiser has to identify the type of value (for fair market value, usually the IRS definition: a price between a willing buyer and a willing seller, neither under compulsion, both reasonably informed), identify where the work actually trades, and use comparable data that is relevant, researched, and verifiable.

Selection follows a hierarchy of similarity. You start tight and loosen only as you have to.

Same artist, first and almost always. For fine art, the comparable has to be by the same artist except in the rare case where an artist is so thinly traded that no sales exist at all. This is the single non-negotiable filter, and it is why "art appreciates 8% a year" tells you almost nothing about a specific work. The artist market is paramount. Get that wrong and no amount of careful adjustment saves you.

Then period, medium, size, and subject. Within the artist's market, the appraiser narrows to the same stylistic period, the same medium and support (oil on canvas is not oil on panel and certainly not a work on paper), a similar size, and a similar subject. A signature motif typically trades at a premium to a minor one. A fully resolved, exhibition-grade work sells above a study or a sketch.

Then market context. Recent sales beat old ones, usually within three to five years for a stable market and one to three years for a fast-moving one. Sales at major international houses are not directly comparable to regional auction results, so the appraiser matches the tier to where the subject would most likely sell. And the transaction has to be arms-length. A charity auction, a forced sale, or an insider deal gets discounted or thrown out.

This maps almost exactly onto how we think about a purchase. Step one is choosing the artist market, because that is what drives the return. Step two is finding the best example at the lowest price inside that market. Comp selection is just the disciplined version of step one, run in reverse: instead of asking what to buy, the appraiser asks what the work in front of them is worth, using the same evidence.

What adjustments do appraisers make to comparable prices?

No two artworks are identical, so a raw comp price is never the answer. The appraiser adjusts each comparable to simulate what it would have sold for if it shared the subject's characteristics. The direction is mechanical. If the comp is inferior to the subject on some trait (smaller, weaker condition, less desirable subject), its price gets adjusted up. If it is superior, its price gets adjusted down.

The adjustment factors are consistent across the literature: date of sale (the market trend since), size, medium, condition, subject, period, provenance and exhibition history, and sale venue. Where the market gives enough data, those adjustments are quantitative, expressed as a percentage. Where it does not, they are qualitative, ranked as superior, similar, or inferior, with the appraiser stating plainly that the difference is directional but not precisely measurable. USPAP requires the reasoning be transparent enough that another appraiser could follow how the number was reached, even if they would land somewhere slightly different.

Here is a worked version, with round numbers, so you can see the machinery. Imagine the subject is a 30 by 36 inch oil on canvas landscape by a well-known modern artist, dated around 1955, in very good condition. You find a strong comp: a 28 by 34 inch oil landscape from the same period that sold at a mid-tier auction in 2023 for $100,000. Two adjustments matter. The comp is slightly smaller, so its price gets nudged up. It sold two years ago, so if the artist's market has firmed since, you apply a small upward time adjustment. You repeat this for four to six comps, watch where the adjusted values cluster, and lean hardest on the comps that needed the fewest and smallest adjustments. If the cluster sits between $95,000 and $115,000 and your two closest comps land near $105,000, that is your fair market value, and your report explains why.

The reconciliation is the judgment, not the arithmetic. A comp that required a 60% adjustment is telling you it was never very comparable to begin with.

How does hedonic regression turn art features into dollar adjustments?

The adjustment percentages an appraiser applies are not pulled from intuition. Where the data supports it, they trace back to hedonic regression, the statistical workhorse of art economics. A hedonic model treats a painting's price as the sum of the implicit prices of its features. You regress the log of price on a work's characteristics (size, medium, signature, period, auction house, sale date, and so on), and the coefficient on each feature tells you how much that feature is worth, holding everything else constant. The sale-date coefficients also produce an art price index, which is why hedonic models are the standard alternative to repeat-sales indices for tracking the market.

The two features with the most studied effects are size and medium.

Size matters, with diminishing returns. Across the literature, the elasticity of price to surface area runs roughly 0.3 to 0.8. In plain terms, doubling a painting's area raises its price by something like 30% to 80%, not 100%. This is why a naive price-per-square-inch calculation overstates value for large works. The right adjustment is curved. If the segment's size elasticity is 0.5 and a comp is 50% larger in area than the subject, the size adjustment is about 22% downward, because 1.5 raised to the 0.5 power is roughly 1.22.

Medium sets a clear hierarchy. Oil on canvas typically sits at the top. Watercolors, gouaches, and drawings commonly trade at discounts of roughly 30% to 70% to an oil by the same artist, and prints, being editioned, often run discounts of 60% or more. So if your subject is an oil and your only good comp is a work on paper, the comp's price gets adjusted up by the relevant medium premium, and that single adjustment can be larger than every other adjustment combined.

The other coefficients are smaller but real. Signed works tend to carry a premium of roughly 10% to 30% over unsigned ones. Strong provenance, a museum exhibition history, or a catalogue raisonné reference each tend to add somewhere in the tens of percent. A top-tier auction house commands a premium of roughly 10% to 30% over a regional one for the same work. A good appraiser rarely runs a fresh regression for a single assignment. They use published estimates as benchmarks to calibrate the adjustments, which is the empirical backbone underneath what otherwise looks like pure expert judgment.

Where does comparable sales analysis break down?

This is the harder question, and there is no perfect answer to it. Comparable sales analysis works beautifully for an actively traded artist with dozens of recent sales across sizes and media. It degrades fast for unique or thinly-traded works, and being honest about that degradation is most of what separates a credible appraisal from a confident-sounding one.

The root problem is that art is illiquid and trades infrequently. Global art sales were roughly $65 billion in 2023 across about 39 million transactions, per the Art Basel and UBS report, but that volume is fragmented across millions of unique objects, and most works by major artists sit in museums or long-term collections and almost never trade. For a thinly-traded artist you may have one or two semi-comparable sales in a decade, sometimes none in the right medium or country. A handful of data points cannot produce a statistically reliable value, and the few comps that exist may be years old, embedding a taste cycle that no longer holds. Time adjustments in that setting are conjecture dressed up as math.

Then there is selection bias, which runs in one direction: up. Auction databases capture works that consignors chose to sell publicly and that houses agreed to take, and owners tend to consign after good news, at or near market peaks. Worse, bought-in lots (works that failed to sell) usually vanish from price databases or appear with no price at all. That is survivorship bias with teeth. A failure to sell at estimate is a powerful negative signal, and it is precisely the signal the database deletes. An appraiser who leans only on realized auction prices, especially in a cooling market with rising buy-in rates, will systematically overstate value. The Center for Art Law's framing is the right one: for unique works, fair market value is intensely fact-specific, and the last auction price is not the answer.

So when comps run thin, appraisers demote comparable sales from a pricing engine to a contextual tool and triangulate. They look to analog artists of similar generation, movement, and critical standing. They pull primary-market intelligence from galleries and dealers, including the asking prices that never reached a database and, more usefully, what actually sold and what sat unsold. Where the market is essentially absent, they fall back on replacement cost as a floor. And the report itself changes character. It leans on narrative, presents a value range instead of a false point estimate, and discloses the thinness of the data explicitly.

For an investor, that is the most important translation in the whole piece. Thin comps are a reason to trust the number less, demand a range, and treat the wide band as the real risk disclosure. A precise-looking valuation built on two stale sales is more dangerous than an honest range built on the same two, because the precision is fictional and the range is not.

What does comps methodology mean for art investors?

The reason any of this matters to an investor is that the appraisal method is also the buying method, run in the opposite direction. If you can build a defensible fair value from comparable sales, then a price meaningfully below that fair value is an arbitrage, and a price above it is a tax you are choosing to pay.

That is more or less how we underwrite a purchase. We get the artist market right first, then we fair-value the specific work off comparable sales and the hedonic adjustments behind them, then we look at the asking price. If our team values a work at a million and we can buy it for $700,000, that spread is the opportunity. If it is asking $2 million, we pass, regardless of how much we like the painting.

Two cautions sit on top of that. The first is the A, B, and C problem. A and B examples of an artist tend to appreciate together, while C examples can stay flat for decades, so a comp set that mixes quality tiers without adjusting for them will mislead you in both directions. The second is the thin-comps trap above. The artists where a discount to fair value looks most tempting are often the ones with the fewest reliable comps, which is exactly where fair value is hardest to pin down. Conviction should scale with the depth of the comparable data, not with the size of the apparent discount.

The Bottom Line

  • Comparable sales analysis is the default method for valuing art. An appraiser selects recent arms-length sales of similar works, adjusts each for the ways it differs from the subject, and reconciles the adjusted prices into a fair market value. USPAP requires the reasoning be transparent and the data verifiable.
  • Comp selection follows a hierarchy: same artist almost always, then period, medium, size, subject, condition, recency, and market tier. The closest comps, needing the fewest adjustments, carry the most weight.
  • Adjustments trace back to hedonic regression. Size has an elasticity of roughly 0.3 to 0.8 (diminishing returns, so doubling the area adds 30% to 80%, not 100%), works on paper run 30% to 70% below oils by the same artist, and signature, provenance, and a top-tier auction house each add in the tens of percent.
  • The method breaks down for unique or thinly-traded works. Sparse, stale comps, plus selection and survivorship bias that delete unsold lots, push valuations upward. Honest appraisers respond with analog artists, dealer intelligence, value ranges, and explicit disclosure of the data's thinness.
  • For investors, the appraisal method is the buying method in reverse. A defensible discount to fair value is the edge, but conviction should scale with how deep the comparable data actually is, not with how large the apparent discount looks.

Sources

  1. Lo Appraisals. "Understanding Appraisal Research Methodology: Market Selection and Valuation Approaches." 2024. https://www.loappraisals.com/chicago-art-appraiser-collector-resources/understanding-appraisal-research-methodology-market-selection-and-valuation-approaches
  2. Appraisal Institute. "General Appraiser Sales Comparison Approach." Appraisal Institute Education. https://www.appraisalinstitute.org/education/search/general-appraiser-sales-comparison-approach
  3. ProEducate. "Comparable Sales Analysis and Adjustments." https://www.proeducate.com/courses/static_files/docs/LA/Post/ComparableAnalysis.pdf
  4. "Hedonic Regression and the Construction of Art Price Indices." SSRN Working Paper, 2019. https://papers.ssrn.com/sol3/Delivery.cfm/SSRN_ID3322843_code281689.pdf?abstractid=3322843
  5. Cornell University eCommons. "Hedonic Pricing Models in Art Markets." https://ecommons.cornell.edu/server/api/core/bitstreams/d4c1fc16-4a10-495b-a75a-010006f449b6/content
  6. Center for Art Law. "Vagaries of Valuation for Collections of Artwork." https://itsartlaw.org/art-law/vagaries-of-valuation-for-collections-of-artwork/
  7. Art Basel and UBS. "The Art Market 2024." Art Basel, 2024. https://www.artbasel.com/about/initiatives/the-art-market
  8. Deloitte. "Art & Finance Report 2023." Deloitte Luxembourg, 2023. https://www.deloitte.com/lu/en/services/financial-advisory/research/art-finance-report.html
  9. Appraisal Buzz. "On Comparable Sales." https://appraisalbuzz.com/on-comparable-sales/