Predicting the outcome of the US presidential election was a popular project for diplomats drafting cables and political commentators penning their columns. But it was also the undertaking of people seeking to make a quick buck.
Indeed, the election result in November was a victory both for Donald Trump and for the more than 100,000 users of online prediction markets, such as Polymarket, Kalshi and PredictIt, willing to make a wager on the winning candidate.
Prediction markets – which allow users to trade financial contracts related to future events and effectively bet on the outcomes – have a good track record at forecasting elections. They have also been successfully used to produce business insights by firms including Google and Goldman Sachs.
What does it mean if the crowd is wiser than the experts? If prediction markets can accurately forecast events of major geopolitical consequence, should we be harnessing their wisdom more?
As it happens, these questions have been asked before, including at the top echelons of foreign policymaking.
A closed prediction market involving analysts from Five Eyes countries, for example, could generate forecasts based on all of the agencies’ intelligence, without requiring full disclosure of information.
In 2001, the United States’ Defence Advanced Research Project Agency (DARPA) launched a program to test if intelligence insights could be drawn from wagers made by traders on indicators such as states’ military activity, economic growth, and even on numbers of western terrorist casualties. However, the program was soon shut down after being excoriated by Congress, with one senator describing the idea as a “fantasy league terror game”.
From 2010-20, thousands of analysts from US intelligence agencies participated in a closed “Intelligence Community Prediction Market” (ICPM) run by the Office of the Director of National Intelligence. A 2018 study of this program, though criticised, found that forecasts generated by the market were “significantly more accurate than the imputed forecasts made by analysts” who had read relevant intelligence reports.
Outside of these examples, however, prediction markets, and other experimental forecasting programs, are only rarely used (at least from what is known publicly) by intelligence agencies and diplomatic services.
Granted, prediction markets are fallible; they failed to predict Trump’s victory in 2016, and even in the most recent election they did not forecast that he would win the popular vote. Online prediction markets such as Kalshi have particular methodological limitations, in part because only a narrow subset of the population uses them.
But if designed according to best-practice, in-house prediction markets can serve a useful purpose for intelligence agencies and diplomatic services.
The value of prediction markets is in aggregating information that is otherwise dispersed throughout a large population into a single measure – the price of the financial contract. For behemoth intelligence communities, they can draw in the views of analysts beyond just those with the job of producing a particular assessment.
Prediction markets also allow participants to express their confidence in a particular outcome (the size of the bet) without disclosing the actual information they possess. Given that intelligence agencies are highly protective of their secrets, and wary of the damage that could be caused by revealing their sources, prediction markets can facilitate better cooperation. A closed prediction market involving analysts from Five Eyes countries, for example, could generate forecasts based on all of the agencies’ intelligence, without requiring full disclosure of information.
As well, prediction markets generate a quantitative rather than qualitative assessment of the likelihood of an event occurring. For time-poor officials, terms such as “probable” and “unlikely”, words of estimative probability as they have come to be described in intelligence assessment, can be more easily misinterpreted than numerical figures. In prediction markets, money really does talk.
Finally, a key advantage of prediction markets – which may in fact partially explain why they have not caught on – is that they upend hierarchies. They challenge participants to put their money where their mouths are, regardless of rank or reputation.
Prediction markets aren’t perfect, but intelligence agencies and diplomatic services aren’t either. The 7 October attack and the 2021 US withdrawal from Afghanistan showed that intelligence failures still happen.
Of course, decisionmakers should not blindly follow advice drawn from any one source, and insights from prediction markets should be weighed against other information and analysis. But in a game as complex and as consequential as geopolitics, why not hedge your bets?
