Predicting Rare Events

In late November 2020, almost four weeks after the US presidential election, the race had been called, a transition had been initiated, and vote totals in a decisive number of states had been certified. And yet, the odds that the incumbent would be reelected were hovering around one-in-eight on PredictIt.

Some observers took this as evidence of a deep flaw in prediction markets as forecasting mechanisms. Nate Silver, to take one prominent example, thought that these prices were insane, driven by dumb money and delusional people.

But then January 6 happened, and we now know that there was an orchestrated attempt to hold on to power that came perilously close to succeeding. It would not surprise me in the slightest if some of the people associated with this effort—people in the prior administration, their legal advisers, and supportive outside groups—had accounts on PredictIt and took positions consistent with their beliefs about what was about to transpire.

In doing so, they made a secretive process a little more transparent. Of course, there were people who raised alarms about the possibility of subversion well before the election, but those were people who feared rather than hoped for a reversal of the electoral outcome. In contrast, prediction markets were feeding us information straight from the horse's mouth.

This explanation of the prices in late November seems more plausible to me than the idea that they were irrational reflections of foolishness and self-deception.

The attempt to reverse the outcome of a certified election was a unique event in modern American history. So was the global financial crisis of 2008-2009, though perhaps we could look to the Great Depression for some parallels. The coronavirus pandemic was not quite unique but one would have to go back a century to see something similar. And as for the effects of climate change, it is unlikely that anything in recorded human history can provide much of a guide.

Predicting such events is much more difficult than forecasting more routine occurrences for which we have reasonably good models and a lot of relevant data. To anticipate rare or unprecedented events one needs to rely on human imagination, and to find some way of adjudicating differently imagined futures.

Prediction markets, however imperfectly, provide a platform for doing precisely this. They continually break down viewpoint homogeneity. If all traders in a market have roughly the same opinion, this will be reflected in the contract price, and those with a different opinion will see a bargain. They will be induced to enter the market, thus shifting the price. In prediction markets, and indeed financial markets more generally, prices at any moment reflect a clash of narratives rather than any kind of consensus. One can speak metaphorically about what the market is “telling us” but this message is just an aggregation or consolidation of a set of mutually inconsistent viewpoints.

Markets are not the only mechanism for harnessing decentralized information, and alternative platforms exist that are more collaborative than competitive. For instance, the pol.is digital tool allows for clusters of opinion on any given issue to be visualized and influenced by public statements; this has been put to effective legislative and policy use in Taiwan. But these alternatives are still at an early stage of development, and it will be a while before they achieve the simplicity, standardization, and wide reach that prediction markets already enjoy.

Among the most serious problems we face as a democracy are affective polarization, ideological segregation, echo chambers, disinformation, and propaganda, all of which contribute to disagreement about basic statements of fact. Prediction markets can serve as a modest antidote to this, by inflicting the cold, hard punishment of monetary loss on those who insist on believing and betting on falsehoods. Instead of shutting down PredictIt, the CFTC should consider removing the monopoly power that the exchange enjoys under the terms of the no-action letter. Let other exchanges operate under comparable conditions, and allow for a broader range of contracts that reference publicly verifiable outcomes related to climate and public health. Let those who absorb and disseminate fake news put money where their mouths are, and let them be publicly and punitively corrected. They may not change their minds, but at least their ability to change the minds of others will be diminished.