Auction Theory

Auction Theory is a subfield of microeconomics, game theory, and mechanism design that examines how bidding rules govern strategic behavior, resource allocation, and price formation under information asymmetry. Initiated by William Vickrey's 1961 analysis of sealed-bid markets and expanded by Robert Wilson and Paul Milgrom (recipients of the 2020 Nobel Memorial Prize in Economic Sciences), auction theory models markets as Bayesian games of incomplete information. Its mathematical insights guide the design of sovereign Treasury debt issuances, electromagnetic spectrum license allocations, wholesale electricity pools, and real-time online advertising exchanges.

Open-Outcry and Sealed-Bid Protocols

Standard single-object auctions are categorized into four benchmark mechanisms:

1. English Auction (Ascending Open-Outcry): The auctioneer begins at a reserve price and incrementally raises the standing price. Bidders openly indicate continued participation until only one bidder remains; the winner claims the item at the price where the last rival dropped out.
2. Dutch Auction (Descending Clock): The auctioneer initializes a price clock at a high level and lowers it continuously until the first bidder stops the clock, claiming the object and paying the stopping price.
3. First-Price Sealed-Bid (FPSB) Auction: Bidders independently submit confidential bids b_i. The bidder submitting the highest bid b₍₁₎ = max_j b_j wins the item and pays their own bid b₍₁₎.
4. Second-Price Sealed-Bid (Vickrey) Auction: Bidders submit confidential bids b_i. The highest bidder wins the item but pays a price equal to the second-highest bid b₍₂₎.

Strategic equivalences connect open and sealed formats. The Dutch auction is strategically isomorphic to the First-Price Sealed-Bid auction because a bidder's choice of stopping price depends solely on their private valuation and beliefs about rivals' maximum bids, with no interim information revealed prior to the terminal bid. Under pure independent private values (IPV)—where each bidder i knows their own valuation v_i with certainty and learns nothing about v_i from others' signals—the English ascending auction is weakly strategically equivalent to the second-price Vickrey auction.

Dominant-Strategy Truth-Telling vs. Bid Shading

In a Vickrey second-price auction, bidding one's true valuation b_i = v_i is a weakly dominant strategy regardless of the number of competitors n, their risk attitudes, or the underlying probability distribution F(v). If bidder i overbids (b_i > v_i), the deviation only alters the outcome when the highest rival bid B_−i = max_j≠i b_j falls in the interval (v_i, b_i), causing bidder i to win at a price B_−i > v_i and incur a net loss. Conversely, underbidding (b_i < v_i) forfeits profitable wins whenever B_−i ∈ (b_i, v_i) without lowering the payment when bidder i still wins.

In a First-Price Sealed-Bid auction, bidding b_i = v_i guarantees zero surplus (v_i − b_i = 0) even upon winning. Rational bidders therefore engage in bid shading, submitting b^*(v_i) < v_i to balance the marginal increase in winning probability against the reduced profit margin upon winning. Assuming n risk-neutral bidders with valuations drawn independently from a cumulative distribution F(v) on [0, v̄], the symmetric Bayesian Nash equilibrium bidding function equals the expected value of the highest rival valuation conditional on it being below v_i:

When valuations are uniformly distributed on [0, 1] (F(v) = v), this integral simplifies to b^*(v_i) = ((n − 1) ÷ n)v_i. As competition intensifies (n → ∞), the shading margin (1 ÷ n)v_i vanishes and the first-price bid converges to the true valuation v_i.

The Revenue Equivalence Theorem

Formulated by William Vickrey (1961) and generalized by Roger Myerson (1981) and John Riley and William Samuelson (1981), the Revenue Equivalence Theorem (RET) states that any two auction mechanisms yield the exact same expected revenue to the seller—and the exact same interim expected payment for every bidder of type v_i—provided four conditions hold:

1. Bidders are risk-neutral.
2. Private valuations v_i are drawn independently and identically (i.i.d.) from a strictly increasing, atomless distribution F(v) on a connected interval [v_min, v_max].
3. In equilibrium, the object is always allocated to the bidder with the highest valuation v₍₁₎ (Pareto efficiency).
4. Any bidder with the lowest possible valuation v_min receives zero expected surplus.

For n bidders with uniform valuations on [0, 1], expected seller revenue in the English, Dutch, First-Price, and Second-Price formats is identical and equals the expectation of the second-order statistic E[v₍₂₎]:

Revenue Maximization and Virtual Valuations

When the seller seeks to maximize expected revenue rather than pure allocative efficiency, Myerson (1981) demonstrated that expected revenue equals the expected virtual valuation ψ(v_i) of the winning bidder:

where f(v) = F′(v) is the probability density function and (1 − F(v_i)) ÷ f(v_i) represents the inverse hazard rate measuring bidder i's informational rent. In a symmetric regular environment where ψ(v) is strictly increasing, the revenue-maximizing auction is a standard auction supplemented by a binding reserve price r^* = ψ⁻¹(c₀), where c₀ is the seller's valuation. For uniform valuations on [0, 1] with c₀ = 0, ψ(v) = 2v − 1, yielding an optimal reserve price r^* = 0.50 independent of the number of bidders n.

Common-Value Auctions and the Winner's Curse

In common-value (or affiliated interdependent-value) auctions—such as offshore oil tract leases studied by Capen, Clapp, and Campbell (1971) and formalized by Robert Wilson (1977)—the item possesses an identical true unknown value V for all bidders, but each bidder observes only a noisy private geological signal s_i = V + ε_i with E[ε_i] = 0. Even though each signal is unbiased unconditionally (E[V | s_i] = s_i), winning the auction selects the bidder with the most optimistic upward estimation error max_j ε_j > 0:

Failure to condition on the adverse informational event of winning produces the Winner's Curse, causing naive bidders to overpay and suffer systematic losses. Rational Bayesian inference requires bidders to shade their bids downward as a rising function of n. Furthermore, Milgrom and Weber's (1982) Linkage Principle establishes that when bidder signals are positively affiliated, open ascending (English) auctions generate higher expected seller revenue than sealed-bid formats because observing rivals' dropout prices reduces informational uncertainty and mitigates defensive bid shading.

Spectrum and Combinatorial Auctions

When selling multiple heterogeneous items with strong complementarities—such as regional radio spectrum licenses whose value depends on assembling contiguous national networks—independent sequential or simultaneous sealed auctions expose bidders to the exposure problem. To resolve this, Paul Milgrom, Robert Wilson, and Preston McAfee engineered the Simultaneous Multiple Round Ascending (SMRA) auction for the U.S. Federal Communications Commission (FCC) in 1994, pairing simultaneous ascending clocks across all licenses with activity rules that prevent late-stage "sniping." Modern spectrum and electricity markets increasingly deploy combinatorial clock auctions and generalized second-price (GSP) mechanisms to price package bids and sponsored search slots.

How does bidder risk aversion break Revenue Equivalence?

When bidders exhibit concave utility functions (risk aversion), they fear losing the object in a First-Price Sealed-Bid auction and consequently shade their bids less, bidding closer to v_i to insure against losing. In contrast, risk aversion does not alter the dominant strategy b^*(v_i) = v_i in a Second-Price or English IPV auction. Thus, under risk aversion, First-Price auctions generate strictly higher expected seller revenue than Second-Price auctions.

Why does Google use Generalized Second-Price (GSP) auctions instead of VCG for search ads?

Although VCG auctions make truthful bidding a dominant strategy across multiple ad slots, Benjamin Edelman, Michael Ostrovsky, and Michael Schwarz (2007) showed that the Generalized Second-Price (GSP) auction—where the k-th highest bidder wins slot k and pays the (k + 1)-th bid per click—was simpler for advertisers to understand and achieves a locally envy-free equilibrium whose revenue matches or exceeds the VCG outcome.

What is an All-Pay Auction and where does it occur in economics?

In an all-pay auction, every participant forfeits their submitted bid b_i regardless of whether they win the prize, while only the highest bidder captures the reward. All-pay auctions model political lobbying, research-and-development patent races, litigation expenditures, and biological contests.

1. Vickrey, W. (1961). Counterspeculation, Auctions, and Competitive Sealed Tenders. The Journal of Finance, 16(1), 8–37. https://doi.org/10.1111/j.1540-6261.1961.tb02789.x
2. Milgrom, P. R., & Weber, R. J. (1982). A Theory of Auctions and Competitive Bidding. Econometrica, 50(5), 1089–1122. https://doi.org/10.2307/1911865
3. Riley, J. G., & Samuelson, W. F. (1981). Optimal Auctions. The American Economic Review, 71(3), 381–392. https://www.jstor.org/stable/1802786
4. Milgrom, P. (2004). Putting Auction Theory to Work. Cambridge University Press. https://doi.org/10.1017/CBO9780511813825
5. Edelman, B., Ostrovsky, M., & Schwarz, M. (2007). Internet Advertising and the Generalized Second-Price Auction: Selling Billions of Dollars Worth of Keywords. American Economic Review, 97(1), 242–259. https://doi.org/10.1257/aer.97.1.242