Art 1: Art market fundamentals

View study sheet (PDF) View SQL cheat sheet (PDF)

This session asks a single plain question and answers it with real transaction data: what forms the price of a work of art. Rather than treating price as a mystery or a matter of taste alone, students learn to read it as the outcome of several factors working together, the artist, the medium and format, the size, the period, the rarity, the provenance, the condition, and the timing of the sale. The dataset records what buyers actually paid at more than 850 vendors with history back to 1949, so students can see the secondary market as it behaves, where a pre-sale estimate meets a real buyer, and where two works by the same artist can sell for very different sums. No finance background and no coding are assumed. The session builds toward a comparable set, the tool every appraiser and dealer uses to judge what a given work should be worth, built here by hand from real sold lots. This is Art 1, the entry point of the Art market track.

Target course(s) and level

Art history, art business, arts administration, or museum studies course. Suitable for undergraduates with no prior exposure to markets or data, graduate students entering the art trade, collectors and continuing-education students, and anyone who wants to understand auction results without a finance or coding background. Teach this before the rest of the Art track (Art 2–5).

Learning objectives

By the end of this session, students will be able to:

1. Name the factors that form an artwork's price, including artist, medium and format, size, period, rarity, provenance, condition, and market timing, and explain how each one moves a price up or down.
2. Distinguish the primary market, where a work first sells through a gallery or dealer, from the secondary market, where a work resells at auction, and state which market this dataset captures.
3. Define pre-sale estimate as the auction house's formalized judgment of comparable value and realized price as what a buyer actually paid, and explain how the two relate.
4. Build a comparables set for a single artist from sold lots, and read it for its typical price, its range, and its outliers.
5. Describe price dispersion within one artist's market, using the minimum, median, and maximum realized price, and explain why two works by the same artist can sell for very different amounts.
6. Identify the limits of auction data for art market fundamentals, including that it captures the secondary market only, that documented fields do not include medium, size, provenance, or condition directly, and that the newest periods are still being ingested and should not be read as a trend.

Prerequisites

No prior art market, finance, or data experience is assumed. An introductory art history course or general interest in how art is bought and sold is enough. No SQL or coding background is required; the session is designed to be followed by students who have never queried a database before.

Materials and access needed

• Sandbox access at sandbox.altfndata.com, self-registered with a work or school email, auto-approved, no API key needed.
• Projector or screen share for the instructor demo.
• The coverage browser tab, used before the demo to confirm how a chosen artist appears in the designer field and how a chosen auction house appears in the vendor field.
• A shared reference sheet of the documented fine art fields (designer, model, item_title, sale_date, usd_price_decimal, sale_estimates_high_usd_price, status, vendor, stock_ticker), handed out or projected for students unfamiliar with data tables.

Session outline (75 minutes)

• 0 to 10 min: What forms a price. Introduce the factors, artist, medium and format, size, period, rarity, provenance, condition, and timing, and explain that this session tests these ideas against real sold lots rather than treating price as unexplainable.
• 10 to 20 min: Sandbox orientation. Confirm students can open the fine art data table and locate designer, vendor, usd_price_decimal, sale_estimates_high_usd_price, and status in the data dictionary, with no query written yet.
• 20 to 35 min: Guided demo, build a comparables set for one artist, ordering sold lots by realized price.
• 35 to 50 min: Guided demo, read price dispersion for the same artist using minimum, median, and maximum realized price, and compare estimate to realized price.
• 50 to 60 min: Guided demo, a rough medium and format proxy using text search on item_title, framed explicitly as an imperfect keyword match rather than a clean field.
• 60 to 70 min: Small-group exercise, students pick a different artist and build their own comparables set and dispersion figures.
• 70 to 75 min: Class discussion and homework assignment, groups share what surprised them about their artist's range.

In-class demo (sandbox-first, no code)

1. Open sandbox.altfndata.com, sign in, and select the fine art data table from the SQL editor dropdown.
2. Open the coverage browser tab and confirm how a well-known artist is written in the designer field. Spelling and partial-name matching both matter for a clean comparables set.
3. Return to the SQL editor and run the first guided query, a comparables set for one artist, which lists that artist's sold lots ordered from highest to lowest realized price.
4. Read the list aloud with the class. Point out that the works clustered near the top and bottom are the outliers, and ask what might explain a lot near the top, a larger size, a rarer subject, a better provenance, versus a lot near the bottom.
5. Run the second guided query, price dispersion for the same artist, returning the minimum, median, and maximum realized price alongside the sold-lot count. Explain that the gap between minimum and maximum is the range a comparables set has to explain, and that the median, not the average, is the safer typical figure when a handful of very high prices are present.
6. Run the third guided query, estimate versus realized price for the same artist's sold lots, and explain that the estimate is the house's formalized judgment of comparable value before the sale, while realized price is what a real buyer decided the work was worth on the day. Point out where the two agree and where they diverge.
7. Explain the secondary-market caveat before anyone treats these figures as the full market: this dataset captures resale at auction only, not the artist's primary gallery or dealer sales, so a young or represented artist may have most of their market activity happen where this data cannot see it.
8. Run the fourth guided query, a medium and format proxy using a text search on item_title for a phrase such as oil on canvas or works on paper, and explain clearly that this is a keyword guess at medium, not a documented field, since the fine art table has no medium, size, provenance, or condition column. Compare the price range for the matched subset against the artist's full range, and explain the recency caveat: the newest periods are still being ingested, so no single recent quarter should be read as a shift in an artist's market.

Datasets and queries used

Dataset: fine art data (documented fields: designer, model, item_title, sale_date, usd_price_decimal, sale_estimates_high_usd_price, status, vendor, stock_ticker). In the fine art data the designer field holds the artist or maker name, and the vendor field holds the auction house. The same queries can be repeated against the works of art data table (all_works_of_art_data), which shares this same field set.

Query 1, comparables set for one artist, sold lots ordered by realized price:

SELECT designer AS artist,
       item_title,
       vendor,
       sale_date,
       usd_price_decimal AS realized_price_usd
FROM all_fine_art_data
WHERE status = 'sold'
  AND designer LIKE '%Warhol%'
ORDER BY usd_price_decimal DESC
LIMIT 50;

Query 2, price dispersion for one artist, minimum, median, and maximum realized price:

SELECT designer AS artist,
       COUNT(*) AS sold_lots,
       MIN(usd_price_decimal) AS min_realized_usd,
       approx_percentile(usd_price_decimal, 0.5) AS median_realized_usd,
       MAX(usd_price_decimal) AS max_realized_usd
FROM all_fine_art_data
WHERE status = 'sold'
  AND designer LIKE '%Warhol%'
GROUP BY designer;

Query 3, estimate versus realized price for the same artist's sold lots:

SELECT item_title,
       sale_date,
       sale_estimates_high_usd_price AS high_estimate_usd,
       usd_price_decimal AS realized_price_usd
FROM all_fine_art_data
WHERE status = 'sold'
  AND designer LIKE '%Warhol%'
  AND sale_estimates_high_usd_price > 0
ORDER BY sale_date DESC
LIMIT 50;

Query 4, a rough medium and format proxy via text search on item_title, compared against the artist's full range:

SELECT COUNT(*) AS matched_lots,
       MIN(usd_price_decimal) AS min_realized_usd,
       approx_percentile(usd_price_decimal, 0.5) AS median_realized_usd,
       MAX(usd_price_decimal) AS max_realized_usd
FROM all_fine_art_data
WHERE status = 'sold'
  AND designer LIKE '%Warhol%'
  AND LOWER(item_title) LIKE '%oil on canvas%';

Discussion questions

1. A comparables set for one artist shows lots ranging from a modest sum to many times that amount. What factors, artist reputation aside, might explain a work near the top of that range compared with one near the bottom?
2. Why is the median a safer measure of an artist's typical price than the average, when a comparables set includes a few very high sales?
3. The pre-sale estimate is the auction house's formalized judgment of comparable value. What does it mean when realized price consistently lands above that judgment, and what does it mean when it consistently lands below?
4. This dataset captures the secondary market only, resales at auction, not the primary market of gallery and dealer sales. What kind of artist's true market would this data most understate, and why?
5. A text search on item_title for a phrase like oil on canvas is a rough proxy for medium, not a clean field. What kinds of works might this search miss or wrongly include, and what would you want to check before trusting the result?
6. Two works by the same artist, similar in subject and date, sell for very different prices. Beyond medium and size, what role might provenance and condition play, even though neither is a field in this data?
7. Why is it unwise to read the newest quarter of an artist's sales as evidence that their market is rising or falling, and what would you look at instead to judge a genuine shift?
8. If you were building a comparables set to advise a client on what a specific work might sell for, what would you want in addition to the fields available here, and how would you flag that gap to the client?

Homework assignment

Each student selects one artist represented in the fine art data and writes a two-page comparables memo, as if preparing a client for an upcoming sale. The memo reports the artist's comparables set of recent sold lots, states the minimum, median, and maximum realized price, and identifies at least two outlier lots with a plausible explanation for why each sits where it does in the range, drawing on medium, size, period, rarity, provenance, or condition as concepts even where the data cannot confirm them directly. The memo compares estimate to realized price for at least three lots and states what the pattern suggests about how the house's judgment held up against real buyers. The memo must include the SQL queries used as an appendix, at least one exported table, and an explicit limitations section addressing the secondary-market-only nature of the data, the imperfect nature of any item_title text search used, and the recency caveat. Grading criteria: correct construction and reading of the comparables set (25 percent), correct computation and interpretation of price dispersion and the estimate-to-realized comparison (30 percent), quality of the reasoning connecting outliers to plausible price-forming factors (30 percent), and honest treatment of the analysis's limitations (15 percent).

Going deeper

Key terms

• Comparables: a set of similar sold works used to judge what a given work should be worth, the core tool behind this session's exercises.
• Realized price: what a buyer actually paid for a lot, recorded in usd_price_decimal, as distinct from any pre-sale estimate.
• Pre-sale estimate: the auction house's formalized judgment of comparable value before a sale, recorded here as sale_estimates_high_usd_price.
• Primary market: the market for a work's first sale, typically through a gallery or dealer, not captured in this dataset.
• Secondary market: the resale market captured by auction data, where a work changes hands after its first sale.
• Price dispersion: the spread between the lowest and highest realized prices within a single artist's market, measured here with minimum, median, and maximum.
• Attribution: the identification of a work as by a particular artist, a factor that can swing price sharply when it is uncertain or contested.
• Provenance: a work's ownership history, a factor known to move price but not present as a field in this data.
• Condition: the physical state of a work, another price-forming factor not present as a field in this data.

Common pitfalls

1. Reading the average realized price for an artist rather than the median, letting a small number of very high sales distort the typical figure.
2. Treating a text search on item_title, such as oil on canvas, as a reliable medium field rather than an imperfect keyword proxy that will miss variant phrasing and catch false matches.
3. Building a comparables set on too few sold lots, producing a range too noisy to support a real judgment about a specific work.
4. Reading the newest quarter's sales as evidence that an artist's market is rising or falling, when the newest periods are still being ingested.
5. Forgetting that this dataset is secondary market only, and drawing a conclusion about an artist's overall market strength without accounting for gallery and dealer activity the data cannot see.
6. Explaining a price outlier with a single factor, such as size alone, when attribution, provenance, condition, and timing are usually acting together.

Additional queries to explore

1. Comparables set for a second artist entirely, to compare how wide or narrow the price dispersion is for a market with heavier trading volume against one with lighter volume.
2. Estimate-to-realized comparison across an artist's full sold-lot history rather than a recent slice, to see whether the house's judgment has tended to run high or low over a longer stretch.
3. The same medium and format text search repeated with a different phrase, such as works on paper or bronze, to see how differently priced the matched subset is compared with the oil on canvas subset for the same artist.

Extension activities

1. Have each small group present their artist's comparables set to the class as a short client-facing summary, explaining the range and the outliers in plain language rather than in query syntax.
2. Ask students to repeat the same four queries against the works of art data table (all_works_of_art_data) for an artist or maker who appears there, and compare how the comparables set and dispersion look in a different category of object.
3. Have students research one artist's known auction highlights outside the dataset, such as a widely reported record sale, and check whether that sale appears in their comparables set, to make the idea of provenance and condition driving a headline price concrete rather than abstract.

Connections to other modules

Next in this track: Art 2, the auction business model, explains why the house sets a pre-sale estimate and how its incentives shape that judgment. Art 3, the business of the art market, builds on the comparables and dispersion habits taught here into league tables and artist-level pricing power. Art 4 covers auction theory; Art 5 takes the same fine art data into research methods (indices, demand signals, cohort comparisons). Cross-track: Finance 2 (consumer and luxury economics) for demand-side brand metrics; Data 2 (visualization) for presenting a comparables set.