Winner's Curse¶
Core Idea¶
The winner's curse is the structural result that, when multiple parties compete for a prize of uncertain common value by submitting noisy private estimates, the act of winning is itself informative bad news: the winner is disproportionately the party who most overestimated the value. The phenomenon was first named and quantified by petroleum engineers Capen, Clapp, and Campbell (1971) studying competitive bidding for offshore oil leases, where they observed that winning operators systematically earned below-market returns despite no errors in their geology. [1] The mechanism is purely a matter of selection on an order statistic: when many independent estimates cluster around a true value, the maximum of those estimates is a biased sample, lying above the true value by a margin that grows with both the number of competitors and the dispersion of estimation error. Because the highest bid wins, the winning estimate is precisely this upward-biased maximum, so the winner systematically overpays unless they shade their bid downward to correct for the conditional information contained in the very event "I won." [1]
What makes the winner's curse a genuine structural prime rather than a folk warning about overpaying is its independence from psychology. It bites even when every bidder is perfectly rational, perfectly calibrated, and identically informed; no overconfidence, greed, or error of judgment is required. The curse arises from the geometry of competitive selection alone, which is why the rational correction — bid not your unconditional estimate of value but your estimate conditioned on the hypothesis that yours is the highest of N estimates — is the canonical remedy and the structural heart of the prime. [2]
How would you explain it like I'm…
Winning means you paid too much
Highest bidder usually overpays
Winner's curse
Structural Signature¶
The winner's curse encodes a structural pattern: common uncertain value → independent noisy estimates → selection on the maximum → upward-biased winner → conditional correction. It separates two quantities that intuition tends to fuse: the unconditional expected value of a prize, and the expected value of that prize conditional on having won the right to it. The gap between them — strictly negative for the winner under common-value competition — is the curse, and the structural insight is that this gap is a deterministic consequence of order-statistic selection, not a contingent feature of any particular auction format. [3]
Recurring features:
- Winning a contested estimate is conditionally bad news
- Selection on the maximum of noisy estimates biases the winner upward
- Conditional expected value given "I won" falls below unconditional value
- Shade your bid to correct for the information in winning
- Overpayment as an order-statistic effect, not a psychological error
- Adverse selection imposed by the structure of competition itself
- The most optimistic estimator wins and therefore most regrets
The signature is robust across the number of competitors: the more rivals bidding against the same common value, the more extreme the winning estimate must be to prevail, and the larger the required downward correction. This counterintuitive scaling — that more competition makes winning worse news, not better — is one of the cleanest diagnostic markers that a situation instantiates the winner's curse rather than some other selection effect. [4]
What It Is Not¶
The winner's curse is not simply "paying too much." Overpayment can arise from impatience, sentiment, strategic deterrence of rivals, or plain miscalculation. The winner's curse names a specific structural source of overpayment: the correlation between winning and overestimating that exists even among flawless reasoners. A bidder who overpays because they fell in love with a house is not suffering the winner's curse; a bidder who overpays because they failed to condition on the informational content of beating nine rivals is. [5]
It is not a claim that winning is always bad. In private-value settings — where each party's valuation is genuinely their own and uncorrelated with others' (you want the painting because you like it, full stop) — there is no common value to overestimate, and winning carries no adverse inference. The curse is specific to common-value or correlated-value environments, where the same underlying quantity (oil in the ground, the target firm's true cash flows, the candidate's actual productivity) determines everyone's payoff. Conflating the two settings is the most common misreading of the prime.
It is not a guarantee of loss, only of conditional bias. A disciplined bidder who correctly shades can win and still profit; the curse describes the penalty for failing to condition, not an inescapable verdict on winners. Nor does the prime assert that the value estimate was unbiased to begin with — it adds a selection-induced bias on top of whatever estimation quality exists. Finally, it is not the same as regret or hindsight: the curse is forecastable ex ante from the structure, whereas regret is an ex post emotional response that may or may not track the structural reality.
Broad Use¶
Auctions and procurement. The founding domain. In common-value auctions — oil and gas leases, spectrum licenses, timber rights, Treasury securities — the winner is the bidder whose appraisal of the shared underlying value ran highest. Sophisticated bidders shade aggressively, and the optimal shade increases with the number of competitors and the uncertainty of the appraisal. [3] Reverse (procurement) auctions invert the sign: the firm that wins a contract by bidding lowest is often the one that most underestimated its own costs, leading to the "winner's curse" of unprofitable contracts and downstream renegotiation.
Mergers and acquisitions. The acquirer who outbids every rival for a target is, by construction, the party who placed the highest valuation on the same common asset — the target's future cash flows. This supplies a structural explanation for the persistent empirical finding that acquirers in competitive takeover contests earn poor post-deal returns, the "diminishing returns to bidding" that corporate-finance scholars have documented across decades of merger waves. [6][7]
Labor markets. When several firms compete for a candidate of uncertain productivity, the firm that wins the bidding war (the highest salary offer) is disproportionately the one that most overrated the hire — and may then suffer the regression-to-the-mean disappointment of a star performer who turns out merely good. Symmetrically, the firm that "wins" by hiring a candidate every other firm rejected faces a different adverse-selection problem.
Online platforms, dating, and matching markets. The phrase "the only one who said yes" captures a winner's-curse-adjacent intuition: in markets where many evaluators screen the same candidate, accepting the option that everyone else declined can signal a hidden defect rather than a lucky find. [8]
Evolutionary biology. Contests for mates or territory that are won by the individual willing to invest the most energy or take the greatest risk can leave winners worse off than non-contestants — a biological analogue in which selection on the most "optimistic" willingness-to-pay produces phyrric victory.
Clarity¶
The prime makes explicit a counterintuitive inference that ordinary reasoning suppresses: conditioning on the event "I won" changes the expected value of the prize. Naive decision-makers treat the win as informationally inert — "I won, so now I own the asset I valued at X" — when in fact the win is a signal that X was probably too high. The clarity the prime supplies is to convert a vague unease about competitive overpayment into a precise question: what does the fact that I won tell me about the value I was estimating? [5]
It also clarifies which situations are dangerous and which are safe. By insisting on the distinction between common value and private value, the prime tells a practitioner exactly when the adverse inference applies. Bidding for a one-of-a-kind item you personally cherish carries no curse; bidding for a financial asset whose value is the same for everyone carries the full curse. This sorting is far more useful than a blanket caution against overpaying, because it identifies the specific informational structure that triggers the effect.
Manages Complexity¶
The winner's curse compresses a sprawling family of phenomena — auction overpayment, merger disappointment, post-hiring regression, procurement cost overruns, adverse selection in matching markets — into a single corrective rule: bid the conditional expectation of value given that you are the high estimator, not the unconditional expectation. [9] One mental operation handles all of these cases, because all of them share the same order-statistic structure. A practitioner who internalizes the rule no longer needs a separate heuristic for each domain; they need only ask how many competitors there are and how uncertain the common value is, and shade accordingly.
This compression also tames the analysis of strategic behavior. Game-theoretically, the equilibrium response to the winner's curse is for every bidder to shade, which means the observed winning bids in a competitive market already embed the collective correction. Understanding this prevents a second-order error — assuming one can profit simply by shading a little when everyone else is shading too. The prime thus organizes both the individual's correction and the market's equilibrium into one coherent picture. [3]
Abstract Reasoning¶
The winner's curse supports clean counterfactual reasoning: What if the winner had been chosen at random rather than by highest estimate? Under random selection the curse vanishes entirely, which isolates the cause — it is the selection rule (pick the maximum) and not the noise itself that creates the bias. This counterfactual is the sharpest tool for diagnosing whether a real situation truly instantiates the prime: if changing the winner-selection rule to something value-independent would eliminate the systematic overpayment, the winner's curse is at work. [2]
The prime also enables transfer-by-analogy across domains that look superficially unrelated. The same conditional-on-selection reasoning explains regression to the mean after any competitive screen (the rookie of the year who slumps in year two), publication bias in science (the studies that clear the significance threshold overstate effect sizes), and the optimizer's curse in decision analysis (the option that scored highest in a noisy evaluation has an overstated true score). Recognizing these as instances of one structure lets an insight earned in auction theory be redeployed in statistics, sports analytics, or research methodology. [10]
Knowledge Transfer¶
The auction-theoretic correction transfers directly to deal-making, hiring, and procurement: in each, "winning a contested selection on a noisy estimate of a shared value" warrants a downward (or, in procurement, upward) adjustment whose size scales with the field of competitors. A corporate-development team that has absorbed the auction literature can apply the identical shading logic to acquisition bids; an analyst who understands regression to the mean can recognize the same structure when a fund's top-performing manager reverts. The vocabulary of "conditioning on winning" gives practitioners in one field a ready handle on problems that specialists in another field discovered first. [11] This transfer is grounded, not merely metaphorical: the order-statistic mathematics is identical across substrates, which is why the same corrective rule is provably optimal in each.
Examples¶
Formal/abstract¶
Common-value sealed-bid auction. N bidders compete for an asset of true common value V (say, the recoverable oil under a tract). Each bidder i privately observes a signal s_i = V + ε_i, where ε_i is mean-zero estimation noise. A naive bidder bids their signal s_i. The winner is whoever drew the largest s_i, which means the winner's signal is the maximum of N draws and therefore lies systematically above V; the expected overestimate grows with N and with the variance of ε. The rational bidder computes E[V | s_i is the highest of N signals], which is strictly below s_i, and bids that conditional expectation instead. The gap between the naive bid and the rational bid is the quantified winner's curse. Mapped back: This abstract setup is the skeleton every applied case shares. Whenever a real situation can be cast as "a shared unknown value, independent noisy appraisals, and a winner chosen by the most extreme appraisal," the conditional-correction remedy applies unchanged, and the magnitude of the required correction is read off from the number of competitors and the noise scale — exactly the two diagnostic parameters the structural signature names.
The optimizer's curse in decision analysis. A decision-maker must choose among K alternatives whose true values are unknown; each is scored by a noisy evaluation, and the option with the highest estimated value is selected. Because selection picks the maximum estimate, the chosen option's estimate is upward-biased: its realized value will, on average, disappoint relative to its score. This is the winner's curse stripped of money and bidding — pure selection on a noisy maximum. Mapped back: The structure is identical to the auction: K alternatives play the role of N bidders, the noisy scores play the role of bids, and "choosing the top-scored option" plays the role of "winning." The remedy is also identical — apply a Bayesian shrinkage that pulls extreme estimates back toward the prior mean before selecting — demonstrating that the prime's corrective logic is substrate-independent and travels wherever a maximum is selected from noisy estimates.
Applied/industry¶
A competitive corporate acquisition. A public company is in play, and five suitors each run independent valuations of its future cash flows, arriving at offers between 1.0 and 1.6 times the prevailing share price. The board accepts the 1.6x bid. By the structure of the contest, the winning acquirer is the one whose model was most optimistic about the same shared cash-flow stream — and is therefore the most likely to have overestimated it. Post-deal, the acquirer books a goodwill write-down as the target's performance regresses toward the consensus most rivals had priced in. A disciplined corporate-development team avoids this by shading its top bid downward in proportion to the number of competing suitors, and by walking away when the contest gets crowded rather than treating heavy competition as validation. Mapped back: The five suitors are the N bidders, the target's true cash flows are the common value V, and the accepted 1.6x offer is the upward-biased maximum signal. The board's acceptance of the highest bid is the selection-on-maximum step, and the goodwill write-down is the realized gap between conditional and unconditional value. The remedy — shade more as competitors multiply — is the auction correction applied verbatim.
Procurement of a fixed-price construction contract. A government agency solicits sealed bids to build a bridge of uncertain true cost C. Six contractors each estimate C with error and bid a price plus margin; the agency awards to the lowest bidder. The winner is the contractor who most underestimated the true cost — the mirror image of the classic curse — and now faces the prospect of completing the job at a loss, triggering change-order disputes, corner-cutting, or default. Experienced contractors protect themselves by padding bids when many rivals are present and by declining to chase contracts in overheated bidding fields. Mapped back: This is the winner's curse with the sign flipped: the common value is a cost rather than a benefit, "winning" means bidding lowest rather than highest, and the bias is downward rather than upward. The structural skeleton — shared uncertain quantity, independent noisy estimates, selection on the extreme, biased winner — is preserved exactly, which is why the same conditional-correction reasoning (shade your estimate toward what winning implies, scaled by the field size) yields the right defensive bid.
Structural Tensions¶
T1: Shading to avoid the curse can cost you prizes worth winning. The conditional correction tells a bidder to lower their bid, but a bidder who shades too aggressively loses auctions they would have profited from, while one who shades too little overpays. The optimal shade is a knife-edge that depends on correctly modeling the number and sophistication of rivals — quantities that are themselves uncertain. The very discipline that protects against overpayment, applied too zealously, becomes a discipline of perpetual losing.
T2: More competition is the signal of both greater opportunity and greater danger. A crowded field can mean the prize is genuinely valuable (many sophisticated parties see worth) or that the eventual winner will be cursed most severely (the maximum of many estimates is most inflated). The same observation — heavy bidding interest — supports opposite inferences, and a decision-maker cannot tell from competition alone which reading dominates without knowing whether the value is common or private.
T3: The correction assumes you know you are in a common-value game, but real situations blend common and private value. Pure common value and pure private value are idealizations; most real auctions mix them (the oil tract is worth a common amount of oil but a private amount depending on your extraction technology). The winner's curse applies only to the common component, and misjudging the common/private split means misjudging how much to shade. A bidder convinced their value is mostly private will shade too little; one who overweights the common component will shade too much.
T4: Equilibrium shading can erase the individual's edge. If every sophisticated bidder shades optimally, the winning price already reflects the collective correction, and no individual can profit merely by being the one who "knows about" the winner's curse. The prime's insight is most valuable precisely where it is least widely held — among naive counterparties — and self-defeating in markets where everyone has internalized it. Knowing the structure does not guarantee an advantage; it only prevents a disadvantage.
T5: Selecting against the curse can select for a different adverse trait. A firm that systematically declines crowded bidding contests to avoid the winner's curse may end up acquiring only assets that no one else wanted — exposing itself to the opposite adverse selection, the "loser's blessing" that is actually a hidden-defect problem. Steering hard away from one selection bias can steer directly into its mirror image, and there is no selection rule that escapes both simultaneously.
T6: The curse is forecastable in expectation but invisible in any single case. The structural bias is a statement about averages over many contests; in a particular auction the winner may have been right and the runners-up wrong. This makes the prime hard to act on under pressure, because a manager who shaded and lost a deal that turned out lucrative will be criticized for timidity, while one who overpaid on a deal that happened to work out will be praised for boldness. The structure is real but its verdict only resolves in aggregate, leaving every individual decision contestable.
Structural–Framed Character¶
Winner's Curse sits at the structural end of the structural–framed spectrum: it is the result that, when multiple parties compete for a prize of uncertain common value by submitting noisy private estimates, the act of winning is itself informative bad news — the winner is disproportionately the party who most overestimated the value. Capen, Clapp, and Campbell named and quantified it studying competitive bidding for offshore oil leases.
The phenomenon is a formal selection-bias result: conditioning on having submitted the highest noisy estimate biases the realized value upward, a fact definable without reference to human practice and carrying no normative charge. The same statistics recur outside markets — a selected estimate is the maximum of many noisy measurements in biology, so the chosen value tends to overstate. Its auction-theory origin, bidding vocabulary, and reliance on competing agents give partial leans, but it is neutral at definition and recognizes the bias already present rather than importing a stance. It reads structural.
Substrate Independence¶
Winner's Curse is a moderately substrate-independent prime — composite 3 / 5 on the substrate-independence scale. Its structure is formally substrate-agnostic — selecting on the highest noisy estimate biases the winning estimate upward, so winning is itself bad news — and it connects cleanly to regression-to-the-mean. The examples reach across auctions, M&A, and labor markets, with a genuine biological analogue in contests won by the most optimistic energy investment and a social-platform case. Yet most uses carry an economic and decision-theoretic flavor, and the pattern does not reach physical or computational substrates structurally, settling it in the middle of the scale.
- Composite substrate independence — 3 / 5
- Domain breadth — 3 / 5
- Structural abstraction — 4 / 5
- Transfer evidence — 3 / 5
Relationships to Other Primes¶
Parents (1) — more general patterns this builds on
-
Winner's Curse is a kind of Adverse Selection
The winner's curse is a specialization of adverse selection in which the self-selection mechanism is a competitive bid for a common-value prize and the selected party is the bidder whose private estimate ran highest. It inherits adverse selection's structure — selection on private information producing a worst-for-the-uninformed-side pool — and specializes by fixing the selection rule to maximum-of-estimates and the asymmetry to noisy common-value valuation. Winning is itself bad news because the winner is disproportionately whoever overshot the true value, an order-statistic fact independent of any error in the bidder's reasoning.
Path to root: Winner's Curse → Adverse Selection → Information Asymmetry → Asymmetry
Neighborhood in Abstraction Space¶
Winner's Curse sits in a sparse region of abstraction space (63rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely rather than landing on a neighbor.
Family — Risk, Arbitrage & Tail Events (14 primes)
Nearest neighbors
- Competition — 0.81
- Adverse Selection — 0.79
- Risk — 0.78
- Bias — 0.77
- Expected Utility — 0.77
Computed from structural-signature embeddings · 2026-05-29
Not to Be Confused With¶
The winner's curse must be distinguished from Randomness, which is its nearest embedding neighbor but a genuinely different prime. Randomness names the presence of noise, chance, or unpredictable variation in a process. It is tempting to read the winner's curse as "just noise causing overpayment," but this misses the structural core entirely. The curse is not the noise itself; it is the selection bias that the act of winning imposes on the noise. Consider the diagnostic counterfactual: take the identical bidders, the identical noisy estimates, and the identical underlying value, but choose the winner at random instead of by highest bid. The randomness is unchanged, yet the curse disappears completely — the randomly chosen winner's estimate is unbiased. This proves that randomness is a necessary ingredient but not the operative cause. The winner's curse requires both noise and a selection rule that conditions on the extreme value of that noise; randomness supplies only the first. Where Randomness is a property of a process (variation exists), the winner's curse is a property of a selection procedure applied to a random process (the maximum is chosen, and the maximum is biased). The two are orthogonal: a process can be highly random with no winner's curse (random selection), and the curse can be arbitrarily severe with only modest randomness if the field of competitors is large enough.
The winner's curse must also be distinguished from Adverse Selection, with which it shares the family resemblance of "selection on hidden value" but differs in mechanism. Adverse selection, in its canonical insurance and used-car forms, is driven by asymmetric information: one party (the seller of the lemon, the high-risk insurance applicant) knows something about the value that the other does not, and the informed party's self-selection into the transaction is what poisons the pool. The buyer ends up with a worse-than-average draw because the better draws were withheld by parties who knew they were better. The winner's curse, by contrast, arises even under symmetric information — when every bidder is equally well-informed and no one possesses private knowledge of the true value. The bias comes not from any party's hidden information but purely from the order-statistic mathematics of selecting the highest of several equally-uncertain estimates. A clean way to see the difference: adverse selection would persist even if there were only one buyer and one informed seller (no competition needed), whereas the winner's curse vanishes if there is only one bidder (no competitive maximum to select). Adverse selection is about who chooses to participate given what they privately know; the winner's curse is about which symmetric participant the competition selects. They can compound — a contested auction for an asset whose seller has private information suffers both — but they are distinct structural sources of disadvantage, and the remedies differ: adverse selection is addressed by screening, signaling, and information disclosure, while the winner's curse is addressed by conditional bid-shading regardless of information disclosure.
Finally, the winner's curse must be distinguished from Overconfidence (and the related optimism bias), which is its most frequently confused neighbor. Overconfidence is a cognitive disposition: a systematic tendency of agents to overrate their own estimates, knowledge, or chances. It would be easy to explain auction overpayment as "bidders are overconfident," and indeed overconfident bidders do overpay. But the winner's curse is fundamentally a structural selection effect that bites even perfectly calibrated, unbiased reasoners. If you populate an auction with bidders whose estimates are individually unbiased — each one's expected estimate equals the true value, no overconfidence whatsoever — the winner's curse is undiminished, because the winner is still selected as the maximum of those unbiased estimates, and the maximum is biased upward even when every individual estimate is not. This is the decisive distinction: overconfidence is a property of individual estimates (each is too high on average), while the winner's curse is a property of the selected estimate (the winning one is too high even when none individually is). Overconfidence and the winner's curse can stack — overconfident bidders in a competitive auction suffer doubly — but the curse is the residual that remains after all overconfidence is purged. Recognizing this prevents the analytical error of attributing to psychology what is actually a consequence of competitive structure, and it explains why the remedy is mathematical (condition on winning) rather than dispositional (just be less confident).
Solution Archetypes¶
No catalogued solution archetypes reference this prime yet.
Notes¶
The winner's curse was named in the petroleum industry but its structural reach is far wider than its origin domain suggests. The single most important practical discipline it implies is to treat winning itself as evidence — to update downward on the value of any prize the moment you learn you have secured it in a contested common-value setting. This runs directly against the psychological grain, since winning feels like confirmation rather than warning, which is part of why the curse persists in markets populated by experienced professionals.
A recurring source of confusion is the boundary with private-value settings. The prime does not apply when valuations are genuinely idiosyncratic, and applying the shading correction there would simply cause a bidder to lose auctions for things they truly want at prices they could afford. The diagnostic question is always whether the prize has a common component whose value is shared across bidders; only that component is subject to the curse.
The procurement (reverse-auction) form deserves explicit flagging because its sign inversion fools practitioners who learned the prime in its classic upward-bidding form. "Winning" by being the cheapest is structurally identical to "winning" by being the most optimistic — both select the extreme estimate of a common quantity — but the resulting bias is a cost underestimate rather than a value overestimate, and the defensive correction is to bid higher, not lower.
Finally, the prime connects to a broad family of selection-on-the-maximum effects — regression to the mean, publication bias, the optimizer's curse — that share the same mathematical skeleton. Treating these as one structure rather than four unrelated cautionary tales is one of the prime's chief contributions to abstract reasoning.
References¶
[1] Capen, E. C., Clapp, R. V., & Campbell, W. M. (1971). Competitive bidding in high-risk situations. Journal of Petroleum Technology, 23(6), 641–653. Seminal paper naming and quantifying the winner's curse from competitive offshore oil-lease bidding, showing the winning bid is the upward-biased maximum of independent noisy appraisals, biased above true value by a margin that grows with competitor count and estimation-error dispersion. ↩
[2] Wilson, Charles. "A Model of Insurance Markets with Incomplete Information." Journal of Economic Theory 16, no. 2 (December 1977): 167–207. DOI: 10.1016/0022-0531(77)90004-7. Foundational analysis of equilibrium in insurance markets with adverse selection, including pooling vs. separating equilibria and the role of participation constraints when the outside option dominates the offered menu. ↩
[3] Milgrom, P. R., & Weber, R. J. (1982). A theory of auctions and competitive bidding. Econometrica, 50(5), 1089–1122. General affiliated-values auction theory establishing that the gap between unconditional value and value conditional on winning is a format-independent order-statistic consequence, that optimal bidding requires shading that rises with competition and uncertainty, and that in equilibrium all bidders shade so winning prices embed the collective correction. ↩
[4] Kagel, J. H., & Levin, D. (1986). The winner's curse and public information in common value auctions. American Economic Review, 76(5), 894–920. Experimental demonstration that adding bidders worsens the winner's curse — large fields (6–7 bidders) bid more aggressively and earn negative profits — confirming that more competitors make winning worse news and require a larger downward correction, a diagnostic marker of the curse. ↩
[5] Thaler, R. H. (1988). Anomalies: The winner's curse. Journal of Economic Perspectives, 2(1), 191–202. Survey establishing the winner's curse as a structural source of overpayment — the correlation between winning and overestimating — distinct from impatience or miscalculation, and making explicit that conditioning on the event "I won" changes the expected value of the prize. ↩
[6] Roll, R. (1986). The hubris hypothesis of corporate takeovers. Journal of Business, 59(2), 197–216. Foundational corporate-finance argument that the acquirer who outbids all rivals is the party with the highest positive valuation error, supplying a structural explanation for systematically poor post-deal acquirer returns in competitive takeover contests. ↩
[7] Boone, A. L., & Mulherin, J. H. (2008). Do auctions induce a winner's curse? New evidence from the corporate takeover market. Journal of Financial Economics, 89(1), 1–19. Empirical takeover-auction study testing the winner's-curse prediction that bidder returns fall with the level of competition and target-value uncertainty, documenting the "diminishing returns to bidding" pattern. ↩
[8] Bazerman, M. H., & Samuelson, W. F. (1983). I won the auction but don't want the prize. Journal of Conflict Resolution, 27(4), 618–634. Classroom common-value (coin/paper-clip jar) experiments showing winners systematically overpay; the winner is disproportionately the most optimistic evaluator, so accepting the option that other evaluators declined signals a likely hidden defect ("the only one who said yes"). ↩
[9] Kagel, J. H., & Levin, D. (2002). Common Value Auctions and the Winner's Curse. Princeton University Press. Comprehensive synthesis of theory and laboratory evidence compressing auction, procurement, and bidding overpayment into the single corrective rule: bid the conditional expectation of value given that you are the high estimator, not the unconditional expectation. ↩
[10] Smith, J. E., & Winkler, R. L. (2006). The optimizer's curse: Skepticism and postdecision surprise in decision analysis. Management Science, 52(3), 311–322. Decision-analytic generalization showing the same selection-on-the-maximum logic produces postdecision disappointment whenever the highest-scored option is chosen from noisy estimates, unifying the winner's curse with regression to the mean and publication bias and prescribing Bayesian shrinkage as the remedy. ↩
[11] Klemperer, P. (1999). Auction theory: A guide to the literature. Journal of Economic Surveys, 13(3), 227–286. Survey of auction theory establishing that the common-value bid-shading correction transfers across applied settings — including procurement, where the sign inverts to a cost-underestimate and the defensive correction is to bid higher — with the required adjustment scaling in the size of the bidding field. ↩