United States Advanced Research Projects Agency

The Defense Advanced Research Projects Agency funded FutureMAP, a proposed program that would have used prediction markets to gather judgments about future events. Its plans included small internal markets for intelligence analysts and a controversial public market focused on the Middle East. The book treats the agency’s role as part of a larger argument: intelligence systems may need ways to aggregate dispersed knowledge, and rejecting an untested method leaves its potential contribution unknown.

FutureMAP and Intelligence Markets

FutureMAP was an abortive plan funded by the Defense Advanced Research Projects Agency. It proposed two kinds of markets through which participants could buy and sell futures contracts based on their expectations about events in the Middle East and elsewhere. The rationale drew on the performance of the Iowa Electronic Markets and other prediction markets: prices could bring scattered judgments together in a form that might reveal information conventional channels missed.

The first component would have been internal markets, limited perhaps to twenty or thirty participants. Intelligence analysts from different agencies—and possibly a small number of outside experts—could trade on the probability of specific events, potentially including terrorist attacks. Because participants could use classified information and hard intelligence, the markets were meant to turn specialized knowledge into collective forecasts without requiring a single superior to decide which analysts’ conclusions deserved attention.

The proposal also addressed the organizational pressures that can shape intelligence judgments. Analysts might feel constrained by internal politics, bureaucratic disputes, or the conclusions their agency or the White House preferred. A market would instead reward accurate forecasting, giving participants a reason to register their assessments even when those assessments were institutionally inconvenient. Its cross-agency membership could also have supplied a form of collective judgment that the intelligence community had struggled to produce.

The Public Market Controversy

FutureMAP’s second component was the Policy Analysis Market, or PAM, which was intended to be open to the public. Like the internal markets, it would have centered on possible developments in the Middle East. Its public accessibility made it politically explosive: critics objected that ordinary people might profit from predictions of catastrophic events.

In the summer of 2003, senators Ron Wyden and Byron Dorgan led efforts to kill PAM, denouncing it as “harebrained,” “offensive,” and “useless.” The idea provoked broad public disapproval among those who learned about it before its cancellation. Wyden dismissed the project as a “fairy tale” and argued that the Defense Advanced Research Projects Agency should spend its money on “real world” intelligence instead.

The book contests the premise that a prediction market would have displaced conventional intelligence gathering. PAM was proposed as an additional way to collect information, not as a replacement for established methods. Opening it to the public might also have drawn out knowledge from people outside the intelligence community—people whose information or perspectives might otherwise remain unheard because of patriotism, fear, or resentment. The argument is not that a public market would necessarily have been accurate, but that its openness need not make it less so: the book’s other market examples show that collective forecasts can be informative even when traders are not specialists.

Aggregation Rather Than Centralization

The proposal belongs to the book’s wider distinction between decentralization and aggregation. Specialized agencies can develop local knowledge and diverse expertise, but those advantages do not automatically combine into a usable whole. In the intelligence community, the problem was not simply that information was decentralized; it was that there was no adequate mechanism for aggregating judgments across the separate agencies. A single central authority might gather information more tightly, but could also sacrifice the diversity, independence, and local knowledge that specialization provides.

Decision markets offered one possible alternative to a traditional top-down structure. Participants could retain access to specialized information while their expectations were brought together through market prices. For the book, that arrangement matters because organizational barriers can prevent useful information from reaching beyond the office or agency where it originates. The market’s proposed advantage was to make individual assessments collectively legible without requiring one central group to sift and rank every piece of evidence.

PAM also raised a moral objection distinct from the question of whether it could forecast well: whether it was wrong to let participants wager on disasters. The book acknowledges the objection’s emotional force but compares prediction with tasks governments already undertake, such as evaluating political stability and the likelihood of catastrophic events. It also notes that intelligence analysts and informants are paid for work that includes anticipating dangers. The controversy therefore concerned not only the market’s practical value, but also whether an unfamiliar public mechanism was more objectionable than comparable forecasting conducted inside government.

Cancellation and the Unanswered Test

The book’s central criticism is that canceling PAM prevented a direct assessment of its usefulness. If the project had produced poor forecasts, its failure could have been established; ending it before it operated ensured that its possible contribution would remain unknown. The proposal was not evidence that prediction markets would solve intelligence failures, but it was an experiment that might have tested whether market incentives could help surface and combine information that conventional institutions missed.

The agency’s role in FutureMAP thus marks a tension between a familiar remedy—putting information under more centralized control—and a more experimental one: preserving specialized sources of knowledge while creating a mechanism to aggregate their judgments. The proposal’s demise left that question unresolved. In the book’s account, the loss was not a proven forecasting tool, but the opportunity to find out whether one could help.