To appear in Alan S. Gerber and Eric M. Patashnik, eds., Promoting the General Welfare: New
Perspectives on Government Performance, Brookings Institution Press, November 2006.
Chapter 7
Decision Markets for Policy Advice
Robin Hanson
The main cause of bad policy decisions is arguably a lack of information. Decisionmakers often do not
make use of relevant information about the consequences of the policies they choose. The problem,
however, is not simply that public officials do not exploit readily available information. It is also that they
do not take full advantage of creative mechanisms that could expand the supply of policy-relevant
information. Among the most innovative and potentially useful information-generating mechanisms are
speculative markets. Speculative markets produce public information about the perceived likelihood of
future events as a natural byproduct of voluntary exchange.
Speculative markets do a remarkable job of aggregating information; in every head-to-head field
comparison made so far, their forecasts have been at least as accurate as those of competing institutions,
such as official government estimates. Many organizations are now trying to take advantage of this effect,
experimenting with the creation of “prediction markets” or “information markets,” to forecast future
events such as product sales and project completion dates.
This chapter examines the uses and limitations of decision markets. Decision markets are
information markets designed to inform a particular policy decision, by directly estimating relevant
consequences of that decision. After reviewing the weaknesses of existing institutions, the mechanics of
decision markets, and a concrete example, the chapter reviews the requirements, advantages, and
disadvantages of decision markets. The chapter also takes a close look at a particular application of this
tool—the controversial yet illuminating attempt to establish a “Policy Analysis Market” to forecast the
consequences of major policy U.S. choices in the Middle East.
Government Information Failures
1