Art 4: Auction theory and behavioral economics

Lesson video · Auction theory and behavioral economics (6:03)

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

Teaching this module? A facilitator guide adds a preparation checklist, timing cues, discussion guidance, slides, and a printable PDF.

This session uses real auction transaction data to test the ideas that auction theory and behavioral economics predict about how buyers and sellers behave in a saleroom. Rather than treating the estimate, the reserve, and the hammer as abstractions from a textbook, students compute pricing power and sell-through directly from more than 15 million transaction records spanning more than 850 vendors, and use those figures to examine three classic ideas together: the estimate as an anchor that shapes what buyers are willing to pay, the winner's curse that can follow from competitive bidding under uncertainty, and the reserve price as the mechanism that produces an unsold lot. The session builds toward a single, testable question: does the data support the textbook predictions, and where does it complicate them.

Target course(s) and level

Behavioral economics, auction theory, or microeconomics elective. Suitable for upper-level undergraduates and graduate students who have completed an introductory microeconomics or game theory course. No finance industry background is assumed.

Learning objectives

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

  1. Define anchoring in the context of a pre-sale estimate and explain why measuring pricing power relative to the high estimate is a direct test of an anchoring effect.
  2. Compute pricing power, the median of realized price over the high estimate, for a category or brand, and interpret a value above or below 1.0.
  3. Explain the winner's curse in a competitive-bidding setting and describe what pattern in the pricing power data would be consistent with it.
  4. Describe the economic function of a reserve price and explain why an unsold lot is a rational outcome for a seller rather than a failure of the auction.
  5. Compute sell-through for a category or house and interpret it as a demand signal that is shaped jointly by buyer appetite and consignment quality.
  6. Identify the limits of using realized-price data alone to test behavioral predictions, including the recency caveat on the newest quarters of data.

Prerequisites

An introductory microeconomics course or a course covering basic game theory concepts (such as common-value auctions or asymmetric information). No prior exposure to SQL, alternative data, or the auction trade is assumed.

Materials and access needed

  • Sandbox access at sandbox.altfndata.com, self-registered with a work or school email, auto-approved.
  • Projector or screen share for the instructor demo.
  • The coverage browser tab, used before the demo to confirm how a chosen category or brand appears in the data dictionary.
  • A short handout summarizing the three theoretical ideas (anchoring, winner's curse, reserve price and sell-through) for reference during the small-group exercise, prepared by the instructor.

Session outline (90 minutes)

  • 0 to 15 min: Introduce the three ideas the session will test: the estimate as an anchor, the winner's curse, and the reserve price as the mechanism behind an unsold lot. Preview pricing power and sell-through as the two measures students will compute.
  • 15 to 25 min: Sandbox orientation. Confirm students can open a data table and locate designer, item_title, sale_date, usd_price_decimal, sale_estimates_high_usd_price, status, and vendor in the data dictionary.
  • 25 to 45 min: Guided demo, compute pricing power for a single category or brand and discuss what a value above 1.0 implies about the estimate as an anchor.
  • 45 to 60 min: Guided demo, compute sell-through for the same category and discuss the reserve price as the mechanism separating a sold lot from an unsold one.
  • 60 to 75 min: Small-group exercise, each group selects a different category or brand, computes both measures, and prepares a short argument about whether the pattern is consistent with anchoring, the winner's curse, or neither.
  • 75 to 85 min: Class discussion, groups report their findings and the class debates where the textbook predictions hold and where the data complicates them.
  • 85 to 90 min: Wrap-up and homework assignment.

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

  1. Open sandbox.altfndata.com, sign in, and select a data table, for example the watches data, from the SQL editor dropdown.
  2. Open the coverage browser tab and confirm how a well-known brand appears in the designer field. Spelling matters for exact filters.
  3. Return to the SQL editor and run the first guided query, pricing power for that brand, and read the result aloud. Ask students what a value above 1.0 implies about the pre-sale estimate as an anchor for bidding behavior.
  4. Introduce the winner's curse and explain that it predicts the eventual buyer in a competitive bid, on average, pays a price that reflects some overestimation of value relative to the field of bidders. Ask students what pattern in pricing power, sustained above 1.0 across many lots rather than a single outlier, would be more consistent with a systematic anchoring or overbidding effect than with noise.
  5. Run the second guided query, sell-through for the same brand, and explain the reserve price as the seller's protection against selling below an acceptable floor. Point out that an unsold lot is retained by the consignor and marked unsold, not discarded from the record.
  6. Explain the recency caveat before anyone reads a single recent period as evidence of a trend: the newest quarters of data are still being ingested, so any time-bucketed comparison should rely on the stable, longer-run pattern rather than the most recent few months.
  7. Open the pre-built charts tab and chart sold-lot count by quarter for the same brand, to show activity over time, and remind students the most recent bars are the least complete.
  8. Ask students, working from the pricing power and sell-through figures alone, to state which of the three theoretical ideas the data for this brand most clearly supports, and which it leaves ambiguous.

Datasets and queries used

Dataset: any of the production category tables (documented fields: designer, model, item_title, sale_date, usd_price_decimal, sale_estimates_high_usd_price, status, vendor, stock_ticker). Examples below use all_watches_data; the same queries run unchanged against all_jewels_gems_data, all_handbags_data, all_fine_art_data, and the other all_*_data tables.

Query 1, pricing power for a single brand (the estimate as anchor):

SELECT designer,
       approx_percentile(usd_price_decimal / sale_estimates_high_usd_price, 0.5) AS pricing_power,
       COUNT(*) AS sold_lots
FROM all_watches_data
WHERE status = 'sold'
  AND sale_estimates_high_usd_price > 0
  AND designer LIKE '%Rolex%'
GROUP BY designer;

Query 2, sell-through for the same brand (reserve price and demand):

SELECT designer,
       COUNT(*) FILTER (WHERE status = 'sold') AS sold_lots,
       COUNT(*) AS offered_lots,
       COUNT(*) FILTER (WHERE status = 'sold') * 1.0 / COUNT(*) AS sell_through
FROM all_watches_data
WHERE designer LIKE '%Rolex%'
GROUP BY designer;

Query 3, pricing power distribution across many brands, used to look for a systematic pattern rather than a single outlier:

SELECT designer,
       approx_percentile(usd_price_decimal / sale_estimates_high_usd_price, 0.5) AS pricing_power,
       COUNT(*) AS sold_lots
FROM all_watches_data
WHERE status = 'sold'
  AND sale_estimates_high_usd_price > 0
GROUP BY designer
HAVING COUNT(*) >= 100
ORDER BY pricing_power DESC
LIMIT 25;

Discussion questions

  1. If pricing power for a brand sits well above 1.0 across hundreds of lots, is that better explained by anchoring on the estimate, by a winner's curse dynamic among competitive bidders, or by the house deliberately setting a conservative estimate? How would you tell these apart with this data alone?
  2. Why is an unsold lot a rational outcome of a reserve price, rather than evidence that the auction mechanism failed?
  3. Sell-through blends buyer demand and consignment quality into a single figure. What additional information would help you separate the two?
  4. The winner's curse is typically described in common-value settings, where bidders share uncertainty about a single true value. How well does that setting describe a luxury auction, where taste and private value plausibly differ across bidders?
  5. If a house consistently sets high estimates relative to eventual realized prices, what incentive might explain that choice, and how would it show up in the pricing power figure?
  6. Why should a single quarter's sell-through figure be read cautiously, given how the dataset is ingested?
  7. What would you expect to see in the data if bidders were not anchored by the estimate at all, and would that pattern be distinguishable from a market where the estimate is simply an accurate forecast?
  8. How might reserve prices differ in their economic function between a category with many undifferentiated lots and a category dominated by a handful of iconic, widely recognized items?

Homework assignment

Each student selects one brand or category represented in any of the production tables and writes a two-page analytical memo applying auction theory to the data. The memo must compute pricing power and sell-through for the chosen brand or category, state explicitly which of the three theoretical ideas (anchoring, winner's curse, reserve price and sell-through) the figures most support, and address at least one alternative explanation for the pattern observed. The memo must include the SQL queries used as an appendix, the exported result set, and an explicit limitations section addressing the recency caveat and the fact that realized-price data alone cannot fully distinguish anchoring from an accurate estimate. Grading criteria: correct computation of pricing power and sell-through (30 percent), soundness of the theoretical argument and consideration of an alternative explanation (35 percent), honest treatment of limitations (20 percent), and clarity of the memo (15 percent).

Going deeper

Key terms

  • Anchoring: a cognitive bias in which an initial figure, here the pre-sale estimate, disproportionately shapes a subsequent judgment or bid.
  • Pricing power: the median of realized price over the high estimate for sold lots, used here as a measurable proxy for anchoring and demand strength.
  • Winner's curse: the tendency for the winning bidder in a competitive, uncertain-value auction to have paid more than the item's average valuation among bidders.
  • Reserve price: the minimum price a seller will accept, below which the lot goes unsold rather than being sold at a loss.
  • Sell-through: the share of offered lots that find a buyer, computed as sold lots divided by all offered lots.
  • Common-value auction: an auction setting in which the item has a single true value that all bidders are uncertain about, the classic setting for the winner's curse.
  • Private-value auction: an auction setting in which each bidder's valuation reflects individual taste and may legitimately differ from other bidders' valuations.
  • Hammer price: the price at which the auctioneer's hammer falls, before any buyer's premium is added, distinct from the total price the buyer pays.

Common pitfalls

  • Treating a single high-profile lot or sale as evidence of a systematic pattern, rather than looking at pricing power across many lots.
  • Reading a low sell-through figure as pure evidence of weak demand without considering that poor consignment quality can produce the same result.
  • Comparing pricing power figures across categories without checking that the sale_estimates_high_usd_price filter excludes zero or missing estimates, which would otherwise distort the ratio.
  • Drawing conclusions about the most recent quarter's activity without accounting for the recency caveat on data still being ingested.

Additional queries to explore

  • Pricing power segmented by house (vendor), to see whether the anchoring pattern differs by which auction house set the estimate.
  • Sell-through trend by quarter for a single category, read as a stable long-run shape rather than a recent-quarter signal, to discuss how demand for a category has behaved over time.
  • Pricing power for unusually high-value lots only (a price threshold), to test whether the anchoring pattern holds as strongly at the top of the market.

Extension activities

  • Have students design a simple experiment, using only the fields available in the data, that would help distinguish an anchoring effect from an accurate pre-sale estimate.
  • Ask students to find a documented real-world auction result, reported in the business press, and compare its pricing power to the category median computed from the sandbox.

Connections to other modules

  • Previous in this track: Art 2–3 for house mechanics and market-level pricing power / sell-through.
  • Next: Art 5, data science for art market research, for research methods on the same measures.
  • Cross-track: Finance 2 for demand-side context behind sell-through and pricing power; Data 3 for a data-driven extension of pattern-finding; Data 6 for a fuller treatment of the recency caveat.