Crypto prediction markets like Polymarket have a cloudy future

US regulators have warned of risks to election integrity and in 2022 fined Polymarket for running an illegal derivatives market.

The ancients used oracles and inspected animal entrails to predict the future. Gen-Z has Polymarket, a crypto platform where (almost) anyone can bet on (almost) anything. It turned out to be the way to pick the winning horse this election year, correctly calling Donald Trump’s clean sweep – unlike most pollsters and pundits  — and making a star out of anonymous trader “Theo,” whose multiple bets netted an estimated $85 million by reportedly focusing on indicators like neighbours’ voting intentions that went unnoticed by a lot of pros.

Yet with regulators in France and the US ramping up scrutiny of Polymarket — including a Federal Bureau of Investigation search that the platform alleged was politically motivated — there’s a need to curb some of the enthusiasm for what some prediction-market proponents see as a “new era” for the industry. And perhaps to wager that if a future shock is coming for the world of opinion polls, it’s more likely to come from artificial intelligence than betting.

The idea behind prediction markets is simple: Bettors’ money flows provide more valuable intelligence than pollsters’ surveys, with financial incentives rewarding the most accurate traders. Like a spread bet on financial markets, the shifting odds can work continuously and also be applied to almost anything, as Polymarket’s morbid array of bets including whether or not a nuclear weapon will detonate in 2024 shows.

But there are limits, and not just because election betting is a legal gray area. One is that these are bets, not polls – they need both closure and liquidity, making them fairly narrow oracles for gauging public opinion. Another is that markets can get it wrong, just like the polls. Two decades ago, George W Bush’s win over John Kerry also seemed to herald a new era for political betting, only for 2016’s double-whammy of Trump and Brexit to shatter it. Market inefficiencies are, of course, there to be exploited by mythical sharp-eyed traders like Theo, but they might also reflect biases inherent to crypto markets (which skew young and male) that diminish their usefulness.

There’s also the risk of manipulation; financial incentives don’t only attract truthseekers. Political scientist Dan Cassino gives the example of centuries-old betting on papal elections in Rome, which led to its own scandals as spreading gossip became lucrative; more recent examples include big bets on Mitt Romney in 2012 that made that year’s election look tighter than it was. US regulators have warned of risks to election integrity and in 2022 fined Polymarket for running an illegal derivatives market. Polymarket’s terms of service describe it as a Panama entity not available in the US, but judging by Bloomberg reports that US-based traders are using the platform, it’s not enough to say caveat bettor.

None of the above absolves the polling industry of its own sins: This will be the third straight election in which experts appeared to understate support for Trump. It’s interesting to note that market-research firm Ipsos SA had coined the term “nouveau nihilism” to describe 2024, based on the fading dream of financial independence and home ownership for younger generations. That ended up encapsulating the final result better than actual surveys.

But we shouldn’t overhype pollsters’ failures either. W. Joseph Campbell, author of Lost in a Gallup, says polls did better this year than in either 2020, which was their worst performance in 40 years, or 2016. There’s been a lot of experimenting with models to reflect Trump support more accurately, he says, and the final error of 1-2 percentage points is a comparatively enviable result. Nonetheless, it’s clear pollsters understated Trump’s support again this year.

The question is now where the prospect of further improvements will come from. Peter Kellner, former chairman of YouGov Plc, notes that traditional polling response rates have slumped along with the use of fixed-line phones, and panels are at the mercy of who will join them. Polling is currently all about modeling, and it may be as good as it can get right now.

One possible answer may lie outside the world of prediction markets, according to Kellner: The realm of artificial intelligence. He points to AI’s ability to crunch huge swathes of data, which could throw up correlations no one has thought of; to build more precise models by tapping into non-demographic data like credit-card swipes, and to analyse existing polls to spot gaps and errors. AI has risks, too, such as the possibility of manipulating public opinion — Microsoft Corp. warned of disinformation risks during the campaign, but it could also bring positives.

This is all in the future, of course. In the meantime, learn to love thy neighbour’s political views – but maybe think twice before placing a crypto bet on them.

Credit: Bloomberg