Major terrorist attacks can have devastating effects on infrastructure, economies, and, most importantly, human life. Modern developments in communications and weapons technologies have supported access by non-state actors to destructive weapons (Shubik, 1997), enabling major terrorist attacks, such as 9/11, the 2002 Bali bombings, and the 2005 London Underground bombings. More recently, the Boston Marathon bombing has acted as a reminder of the continuing threat posed by terrorism. Regardless of the scale, in most cases of terrorism, intelligence agencies fail to identify key information that was available prior to the attack and could have aided in preventing it (Negroponte & Wittenstein, 2010). What more can governments—who already spend billions on national security—do to help prevent terrorist attacks?
One option available to governments is to set up prediction markets on terrorist attacks. Prediction markets (PMs) are marketplaces in which traders can trade shares in predictions of real-world outcomes, such as political, economic, or social events (Weijers, 2013). For example, a prediction market may allow a trader to buy or sell shares in the prediction that ‘A Republican candidate will win the 2016 US presidential election.’ The PM will then pay out a set amount, say $1, for each share held by a trader if that prediction is realised, and $0 for each share held in an outcome which is not realised. The share price represents an aggregate of investor’s perceived likelihood of the prediction being realised, such that if the share price was 55c in a prediction that paid $1 per realised prediction, the aggregate perceived likelihood of the prediction being realised would be 55%.
In the wake of 9/11, the Defense Advanced Research Projects Agency proposed such a solution in the form of the Policy Analysis Market (PAM) (Hanson, 2007). PAM was intended to operate as a public market in which traders could invest real money on political and economic variables, such as US GDP, as well as specific events, such as assassinations and military attacks.1 DARPA’s interest in prediction markets followed the success of the Iowa Electronic Markets, which were launched in 1988 and had been more accurate than traditional forecasting methods, such as polls, in predicting the outcomes of national elections (Berg, Nelson, & Rietz, 2008). However, the reliability of prediction markets as an anti-terrorism intelligence-gathering tool remains unknown. Building up to the trial of PAM, Senators Byron Dorgan and Ron Wyden heard about the project and publicly decried it as “horribly offensive” and a “federal betting parlour on atrocities and terrorism” (Wyden & Dorgan, 2003). As a result, the program was shut down two days later. The comments made by Dorgan and Wyden brought PAM to the attention of politicians and the media, and the repugnance expressed by these groups about PAM has been widely identified as the main justification for the withdrawal of PAM’s funding.2
Why was this kind of market, and these sorts of transactions, thought to be so repugnant? Much of the initial reaction consisted of claims that the program just was “morally wrong,” “disgusting,” “offensive,” “grotesque,” “very sick,” and “morally reprehensible.”3 However, some justifications were given. Robin Hanson (2006)—an economist involved in the development of PAM—has identified three main concerns: first, that PAM would rely on citizens who, due to their lack of training, would be unable to competently undertake or supplement the work of trained intelligence officials; second, that terrorists could mislead intelligence agencies through the market; and third, that terrorists would be able to profit from the market.