Mechanism Design & Auction Theory

The 2020 Nobel in Economics went jointly to Paul Milgrom and Robert Wilson of Stanford for the auction theory and practical formats they had built since 1994 — the FCC spectrum auctions that by then had raised over $200 billion for the US Treasury while laying the institutional foundation for the mobile-telecommunications industry. Spectrum allocation had stalled in the 1980s because licenses are not independent goods: a national carrier needs adjacent regional licenses across different frequency bands, with competitors' moves on every license mattering simultaneously. Milgrom's simultaneous multiple-round auction (1994) let bidders adjust across lots in successive rounds, discovering prices and resolving complementarities iteratively.

Ordinary game theory takes the rules as fixed and asks how clever players will behave. Mechanism design runs the question backwards: fix the outcome you want, then ask what rules would lead self-interested players to produce it. The designer's difficulty is that the information he needs — what a bidder is really willing to pay, which school a family truly prefers — sits inside people with every reason to misreport it. The trick is to make honesty the selfish choice. Take a sealed-bid auction in which the highest bidder wins but pays the second-highest bid, and work out what shading your bid could do for you. Bidding below your true value never lowers the price you pay, because someone else's bid sets that price; it only risks losing an item you would have been glad to win. Bidding above your value can only win you something at a price you did not want to pay. Truth-telling is left as the dominant strategy — and the designer arranged this while knowing nothing whatever about the bidders. The principle generalises: charge each participant not what they offered but the cost their presence imposes on everyone else, and truthful reporting survives into settings with many items. It is elegant and fragile — the same design can be picked apart by bidders who collude, and working out who should win can become computationally intractable. A related problem arises where prices are not permitted at all. Doctors cannot bid for hospital posts and patients cannot buy kidneys, yet both must be allocated. Deferred acceptance solves it by having one side propose while the other holds its best offer so far, releasing it only when something better arrives. The process always terminates, and it terminates in a matching that no pair would both want to abandon. The same algorithm now runs medical residencies and school admissions — and, in its most consequential form, chains of kidney donors, where one incompatible pair is matched with another and both receive a transplant.