Policy Analysis Market

The Policy Analysis Market (PAM) was the proposed public-facing component of FutureMAP, a plan to use prediction markets to improve intelligence about the Middle East and elsewhere. Its cancellation in 2003 left a pointed question unresolved: whether a market could have helped gather dispersed information and resist institutional pressure, or whether the public and political objections to wagering on catastrophe outweighed that potential. The proposal did not replace conventional intelligence gathering; it offered an additional way to collect and aggregate judgments. But it was never tested, so its promised value—and its limits—remained hypothetical.

The intelligence problem behind the proposal

The case for new ways to combine information emerged from a broader concern about how intelligence agencies handled what they knew. Congressional criticism after the failures preceding the September 11 attacks emphasized inadequate information sharing. Rather than assemble a coherent picture of threats, agencies held localized snapshots; Senator Richard Shelby criticized the FBI’s decentralized structure for leaving information fragmented among largely independent fiefdoms. The proposed remedy in that argument was greater central authority. The book challenges that conclusion: when information is dispersed among people and institutions, the task is not necessarily to abolish their independence, but to find a way to make their separate knowledge collectively useful.

That problem has two sides. Decentralization encourages specialization and allows people close to a problem to acquire local knowledge, but it cannot guarantee that discoveries in one part of a system will reach the rest. Effective aggregation must connect those advantages without assuming that a small group of senior officials can simply absorb and interpret everything. The FutureMAP proposal addressed this tension through markets, where participants could register judgments in a form intended to bring otherwise separate information together.

FutureMAP’s two markets

FutureMAP, funded by the Defense Advanced Research Projects Agency, had two proposed components. One was a set of internal markets, expected to be small—perhaps twenty or thirty participants—and open to intelligence analysts and possibly a limited number of outside experts. Those markets might have been used to estimate the probability of particular events, including terrorist attacks. Their participants could draw on classified information and hard intelligence data unavailable to the general public.

The internal-market design was intended to reduce the influence of bureaucratic rivalry and political pressure on analysts’ conclusions. Within agencies, an analyst might feel constrained by what a superior, an institution, or the White House would welcome. A market would instead reward the most accurate forecast, regardless of whether it pleased a boss or helped an agency’s standing. Because analysts from different agencies could participate, the proposed system might also have combined judgments that established structures kept apart. This was a rationale for the design, not a demonstrated result: the markets remained proposals rather than tested intelligence tools.

PAM was the second, more far-reaching component. Unlike the restricted internal markets, it was to be open to the public. Its basic premise was that a market focused on the Middle East could produce useful intelligence, just as the Iowa Electronic Markets and other prediction markets had been used to forecast elections and other outcomes. Opening participation beyond government agencies might also bring in people whose knowledge or observations would not ordinarily reach intelligence officials, whether because of patriotism, fear, or resentment.

Political opposition and cancellation

In the summer of 2003, PAM became the focus of fierce political criticism. Senators Ron Wyden and Byron Dorgan led the effort to stop it, condemning the plan as “harebrained,” “offensive,” and “useless.” The public response described in the book was similarly hostile. A particular objection was that ordinary people might profit from terrible events occurring. The concern was not only whether a market could forecast accurately; to its critics, making catastrophe the subject of wagers was itself morally objectionable.

The proposal’s defenders distinguished prediction from action. PAM was not presented as a substitute for traditional intelligence gathering, but as another channel for collecting information. Its advocates argued that public participation would not automatically make forecasts less accurate: collective judgment can be useful even when traders are not specialists, and a broader pool might include perspectives government agencies would miss. They also argued that market participants would have a direct incentive to make accurate predictions rather than tailor assessments to bureaucratic or political expectations.

The disagreement therefore concerned both institutional design and public values. Critics treated the prospect of wagering on potential catastrophes as offensive and morally wrong. The book acknowledges the visceral unease of betting on an assassination attempt, but points out that government analysts routinely assess questions such as the stability of Jordan’s government, the possibility that the House of Saud might fall, or who would lead the Palestinian Authority in 2005. It further contrasts the market’s proposed incentives with existing practices: intelligence analysts are paid to anticipate dangers, and informants may be paid for useful information. The comparison does not erase the ethical discomfort; it shows why the proposal’s defenders considered forecasting catastrophe a practical intelligence function rather than a unique moral transgression.

What cancellation left unknown

Senator Wyden dismissed PAM as a “fairy tale” and argued that the agency’s resources should go to “real world” intelligence. The book rejects the implied choice between a market and conventional intelligence: no one proposed replacing existing collection with PAM. If the market proved ineffective, a trial could have established that. Canceling it in advance made no such assessment possible.

The argument for PAM also rested on the problem Shelby had identified: information held in separate institutional compartments may never become a useful whole. Markets, in this account, offered a possible way to aggregate judgments across those barriers while giving participants incentives to report what they believed rather than what their organizations preferred to hear. But the proposal’s cancellation means those advantages remain claims about what the market might have done, not evidence of what it actually accomplished.

PAM thus stands at the intersection of two unresolved questions: how to combine dispersed knowledge without simply concentrating authority, and how to judge a forecasting tool whose subject matter provokes moral opposition. Its cancellation did not establish that the market would have improved intelligence, nor did it show that it would have failed. It ensured only that its contribution would remain unknown.