Putting Auction Theory to Work
by Paul Milgrom.
Getting to Work
The era of putting auction theory to work began in 1993-94, with the design and
operation of the radio spectrum auctions in the United States. Although the economic
theory of auctions had its beginnings in the early 1960s, early research had little influence
on practice. Since 1994, economic auction theorists have designed spectrum sales for
countries on six continents, electric power auctions in the US and Europe, CO
auctions, and various asset auctions. By 1996, auction theory had become so influential
that its principal founder, William Vickrey, was awarded a Nobel Prize in economic
featured the success of the US spectrum auctions to justify its support for fundamental
research in subjects like game theory. By the end of 2001, just seven years after the first
of the large modern auctions, the theorists& designs had powered worldwide sales totaling
more than $100 billion. The early US spectrum auctions had evolved into a world
standard, with their major features expressed in all the new designs.
It would be hard to exaggerate how unlikely these developments seemed in 1993.
Then, as now, the status of game theory within economics was a hotly debated topic.
Auction theory, which generated its main predictions by treating auctions as ±games,²
had inherited the controversy. At the 1985 World Congress of the Econometric Society, a
wide gulf developed between bargaining theorists, who were skeptical that game theory
could explain much about bargaining or be useful for improving bargaining protocols,
and researchers in auctions and industrial organization, who believed that game theory
was illuminating their fields. Although game theory gained increasing prominence
throughout the 1980s and had begun to influence the leading graduate textbooks by the
early 1990s, there was certainly no consensus about it in 1994, when the Federal
Communications Commission conducted the first of the new spectrum auctions.
The traditional foundations of game theory incorporate stark assumptions about the
rationality of the players and the accuracy of their expectations that are hard to reconcile
with reality. Economic experimenters have tested the predictions of auction theory in
laboratory experiments with human bidders and found many violations, but some key
tendencies predicted by the theory do find experimental support. The findings indicate
that existing theories oversimplify the way humans play games, and that real world
auction design must be undertaken like other practical arts, by mixing theory with
experiments and practical judgment.
Whatever the doubts in the academy, the dramatic case histories of the new auctions