Most art valuations are looking backward, and that is the single most important thing to understand before you trust any number attached to a painting. An appraisal is an informed estimate of what a work would fetch today, built mostly from comparable sales that already happened, sometimes months or years ago. When the market moves, the appraisal follows with a delay. The practical consequence is that any index or portfolio mark built on appraisals reports returns that look smoother and steadier than the real market underneath. Volatility looks lower than it is. Correlation to stocks looks lower than it is. This is the same problem that quietly inflates the apparent Sharpe ratios of private real estate, private equity, and a good chunk of the hedge fund world. It is worth being precise about, because once you see it, you read every art performance chart differently.

We have spent a lot of time on this question, because it is the reason we build our index the way we do. The short version: there are two honest ways to measure a market this illiquid, and the one most people instinctively reach for is the one that lies to you about risk.

What is appraisal lag and why does it smooth out volatility?

Start with the mechanism, because it is simple and it explains almost everything that follows. An appraiser, or a fund marking an illiquid asset, does not have a live price. So the value gets set as a blend of current evidence and the last known value. The academic version of this is the partial-adjustment model that David Geltner formalized for real estate in 1989 and 1991: the reported return in any period is a weighted average of the true market return and the prior reported value. The valuation only partially catches up to reality each period, then partially catches up again the next period, and so on.

That partial catch-up does two things, and both matter for an investor.

First, it manufactures positive autocorrelation. A move in the real market gets spread across several reporting periods instead of landing all at once. Second, and this is the consequence that should concern you, it shrinks measured volatility. You are effectively running the true price series through a filter that strips out the high-frequency moves. Under Geltner's model, if appraisals adjust only halfway to the true value each period, the reported volatility comes out to roughly 40% of the real volatility. The market could be moving twice as hard as the chart suggests.

A note on how we know this is not just theory. The same math was extended to hedge funds by Getmansky, Lo, and Makarov in their 2004 paper, which modeled reported fund returns as a moving average of true economic returns and showed that illiquid strategies report understated volatility and inflated Sharpe ratios as a direct result. It shows up empirically in real estate too. Desmoothing exercises on appraisal-based indices in the NCREIF style routinely roughly double the volatility estimate, taking annualized figures from the 6% to 8% range up toward 12% to 15% once the serial correlation is stripped out. Same asset. Same period. The risk was always there. The appraisals just hid it.

How is a repeat-sale art index different from an appraisal-based one?

This is where the choice of methodology stops being academic and starts mattering to your returns. There are broadly three ways to build an art index, and they fail in different directions.

Average-price indices track the mean or median price of what sold in a period. They are simple and they are close to useless for measuring appreciation, because the mix of what sells changes constantly. A quarter heavy on Basquiat looks like a rising market even if every individual work is flat.

Hedonic indices model price as a function of an artwork's observable traits, size, medium, artist, date, and a time effect. They use all sold works, which is their strength, but they depend entirely on the model picking up the things that actually drive value, and a lot of what drives art value is not in the spreadsheet.

Repeat-sale indices track the same object across two or more sales and attribute the price change to the market. This is the method Robert Shiller and his collaborators pioneered for home prices, the foundation of the Case-Shiller index, and it is the approach behind Mei Moses and behind our own index. When the same Warhol sells in 2018 and again in 2025, the change in price tells you something real about the market, because you have controlled for the one thing that is hardest to control in art: the specific quality of the specific object.

The reason this matters for the topic at hand: a repeat-sale index is anchored to actual transactions, two real prices a buyer and seller agreed on. An appraisal-based index is anchored to opinions about prices. The first can be noisy and sparse. The second is systematically smooth and systematically late. We will take noisy and honest over smooth and late every time, because you can quantify and correct for noise, and a smooth series actively misleads you about how much risk you are carrying.

What are the biases in a repeat-sale art index?

We are not going to pretend the method we use is clean. It is the right tool, and it still has real biases an honest analyst has to name.

Sample selection. Only works that sell at least twice enter the index. That is a self-selected slice of the market, tilted toward works desirable enough to come back to auction. One survey of the art market noted that works failing to sell can exceed 50% of those offered, and those never enter a repeat-sale sample at all.

Illiquidity and long intervals. Art trades infrequently. A lot of these paintings sit in families for decades, sometimes generations. That is the death, divorce, and debt pattern that drives the market's low turnover, and it means repeat-sale pairs are often years apart. Long gaps make the index sparser and the short-run readings less precise.

Survivorship. The method over-represents works that stayed desirable enough to resell, and quietly drops the ones that fell out of fashion. The Mei Moses index, for context, was built on more than 80,000 works by over 10,000 artists with data going back roughly 200 years, and even that is concentrated in higher-value works and does not represent the whole market.

Here is the honest comparison. Appraisal-based methods bias you toward false calm, understated volatility, and a portfolio that looks lower-risk than it is. Repeat-sale methods bias you toward sparse data and selection effects, problems you can measure, disclose, and partially correct. Neither is perfect. One of them is correctable.

Why does this matter for measuring volatility and correlation?

Because the entire case for owning art as a portfolio asset runs through correlation, and appraisal lag attacks the correlation number specifically.

Think about what happens when a real market move gets smeared across several reporting periods. The part of art's true co-movement with equities that should show up this quarter gets pushed into next quarter and the quarter after. So when you run a simple correlation of art against the S&P using contemporaneous quarterly returns, you get a number that is lower than the true economic relationship. The diversification looks better than it is, for a reason that has nothing to do with diversification and everything to do with stale pricing.

We have always argued that art's low correlation to equities is real and is the point of owning it. True diversification means owning an asset that is largely indifferent to the forces driving everything else. But we want that claim to rest on transaction data that can survive scrutiny, not on an appraisal series whose low correlation is partly a measurement artifact. If your diversification edge disappears the moment someone desmooths your index, it was never an edge. It was an accounting choice.

This is exactly the critique Cliff Asness at AQR has aimed at private markets, the practice he calls volatility laundering. His argument, which the 2024 and 2025 allocator commentary has thoroughly absorbed, is that private equity and private real estate report smooth returns because they are marked infrequently with smoothing assumptions, and that once you unsmooth the series, volatility roughly doubles for real estate, correlation to public markets jumps from the 0.2 to 0.3 range toward 0.6 to 0.7, and the apparent diversification benefit shrinks. The market kept telling on itself in plain numbers. McKinsey's 2025 private markets review reported that buyout secondaries traded at roughly 94% of NAV in 2024, with private markets broadly around 89% of NAV. Buyers were demanding a discount to the reported marks because they did not fully believe them. That discount is the market's own estimate of how stale the appraisals are.

The lesson for art is direct. If you only ever see your art exposure through an appraisal lens, you will systematically underestimate how much risk it carries and overestimate how much it diversifies you. The volatility is not absent. It is deferred.

How should an investor read art performance data with this in mind?

A few practical habits.

Ask what the number is built from. If a return series is based on appraisals or marks rather than repeat transactions, assume the reported volatility is understated, probably meaningfully, and that the true correlation to your equity book is higher than the chart shows. Treat low reported volatility in any illiquid asset, art included, as a flag to investigate the methodology, not as a feature to celebrate.

Watch for too-smooth lines. A return path with almost no quarter-to-quarter variation in an asset that trades a few times a decade is telling you about the valuation process, not the asset. Real illiquid markets are lumpy. Honest indices of them look lumpy.

Size the allocation for the real risk, not the reported risk. We think most investors should hold some art, whether that is 2% or 10%, and we think of it as a long-term, illiquid allocation measured in years rather than quarters. That view does not change because of appraisal lag. What changes is the humility you bring to the volatility figure. Plan around the unsmoothed risk, because that is the risk you actually own.

We hold ourselves to this standard. Our index uses the repeat-sale methodology for exactly the reason laid out here, and we would rather report a number that is harder to produce and honest than one that is easy and flattering. The whole reason this asset class lacked a reliable indicator for so long is that the easy measures were the misleading ones.

Bottom Line

Appraisals look backward by construction, and any performance number built on them inherits that lag. The lag smooths returns, which mechanically understates volatility and understates correlation to public markets, the same flaw that inflates reported Sharpe ratios across private real estate, private equity, and illiquid hedge fund strategies. A repeat-sale index, anchored to two real transaction prices, trades that false calm for sparse but honest data, biases you can measure rather than biases that mislead you. For an investor, the takeaway is to distrust smooth lines in illiquid assets, ask what every return figure is built from, and size your art allocation for the real risk underneath the appraisal, because that risk does not disappear when it goes unmeasured. It just waits.

Sources

  1. Geltner, D. (1989, 1991), appraisal smoothing and partial-adjustment models for real estate returns. Summarized in "Correction Procedures for Appraisal-Based Real Estate Indices," European Real Estate Society. https://eres.architexturez.net/system/files/274_0.pdf
  2. Getmansky, M., Lo, A., and Makarov, I. (2004), "An Econometric Model of Serial Correlation and Illiquidity in Hedge Fund Returns." Referenced via AnalystPrep FRM study notes on illiquid assets. https://analystprep.com/study-notes/frm/illiquid-assets/
  3. "Measuring Systematic Biases in Real Estate Returns," CAIA, 2012, on stale pricing and non-synchronous valuation. https://www.caia.org/sites/default/files/4aiar-measure-2012-q3.pdf
  4. "Illiquidity and Pricing Biases in the Real Estate Market," USC Lusk Center, 2006. https://lusk.usc.edu/sites/default/files/working_papers/wp-2006-1001.pdf
  5. "Demystifying Art Indices," Morgan Stanley, February 2026, on repeat-sale methodology for unique objects. https://www.morganstanley.com/articles/art-market-indexes
  6. "Art Market Performance Index," Art Market Research white paper, October 2025, on repeat-sales regression and its criticisms. https://artmarketresearch.com/wp-content/uploads/2025/10/Art-Market-Performance-Index_White-Paper-2.pdf
  7. "Art market indexes," TIAS Business School, on repeat-sale methodology and the Mei Moses index scope. https://www.tias.edu/en/artikelen/art-market-indexes
  8. "Art Price Indices: Op Ed," Center for Art Law, on selection and unsold-work bias in art indices. https://itsartlaw.org/art-law/art-price-indices-op-ed/
  9. "Global Private Markets Report 2025: Braced for Shifting Weather," McKinsey, on secondaries pricing as a percentage of NAV and PE NAV behavior. https://www.mckinsey.com/industries/private-equity-and-principal-investors/our-insights/mckinseys-global-private-markets-report
  10. "Portfolio Construction Vol. III," Hamilton Lane, October 2024, on appraisal valuations changing more slowly than traded equity valuations. https://www.hamiltonlane.com/en-us/insight/portfolio-construction-vol-3

This material is for general educational and informational purposes only and is not investment advice or a recommendation to buy or sell any security. Past performance and historical index data are not indicative of future results. Index methodologies and the limitations described above affect how performance should be interpreted. Investing in art involves risk, including illiquidity and loss of principal.