Published in Svenska Dagbladet, 2024-10-29. Translated from Swedish.
Want to bet on who will win the US presidential election? With the right prediction you can now make a lot of money. But there are also reasons beyond profit that make speculators want to project a clear winner well ahead of the result.
“People overestimate their knowledge and underestimate the probability that they are wrong.” Nassim Nicholas Taleb, the economist and author whose glass is usually half empty, reviews humanity’s view of itself in his book Fooled by Randomness. Probabilities and truths are tricky. Sometimes they resemble each other, but they are not identical. Now — in the intensifying run-up to the American presidential election — that reminder may be timely. While Americans are placing bets on who will become the next president, the process risks sowing the seeds of distrust about how the election actually played out.
So-called prediction markets have existed for a long time — allowing individuals to speculate on specific questions like whether a certain technology will win a Nobel Prize or whether Taylor Swift will be the most-streamed artist on Spotify this year. It has been a hobby for a small minority trying to foresee the near future. But interest has rarely been higher than now. The reason is a newly legalised and rapidly growing market for wagering on the presidential election result specifically.
In early October the American service Kalshi won its appeal against the CFTC — the Commodity Futures Trading Commission, which regulates these markets. The CFTC had argued that this kind of prediction-making could negatively influence the election, but failed to prove that case sufficiently in court. As a result, Kalshi became the first service permitted to open public, money-backed predictions on who will be the next US president. The brokerage app Robinhood has since also allowed users to bet on the election outcome.
“Prediction” is a generous term in this context. Formally that is what users are supposed to be doing — but since there is a financial interest in the outcome, it is perhaps easier to call it what it actually is: a bet. Even a football pools coupon is a form of prediction, at some level. The difference from ordinary sports betting odds is that the market is entirely open. Kalshi sets no odds of its own; the system functions more like a stock exchange driven by supply and demand. The more people who have bet on a given presidential winner, the lower the payout — because the market has determined that outcome to be more likely. But probability, as noted, is not identical to truth.
The gap between prediction markets and opinion polls is in fact very large. At the time of writing, Donald Trump leads with around 62 percent on Kalshi’s market, while the latest polls show a near-dead heat. At least one of them is very wrong. Which one? And more interestingly, why? Outside the United States, additional prediction markets allow wagering on the election. Polymarket, a crypto-based service, can sidestep American regulations by allowing everyone except Americans to speculate on the result. Foreign users appear to have drawn similar conclusions — Trump leads with around 67 percent on Polymarket.
Some speculators appear extraordinarily confident. A user going by “Fredi9999” recently placed around 150 million kronor on Trump becoming the next president. The underlying crypto technology makes it difficult to know who — or how many people — are behind the position. One can only speculate about the motive. Perhaps it is not primarily about trying to profit from a correct guess, but about trying to influence the outcome. If Trump appears to be the winner in voters’ eyes, his chances of actually winning increase. Rajiv Sethi, economics professor at Barnard College, told the Wall Street Journal: “If I were trying to manipulate a market, this is exactly how I would do it.”
The divergence between prediction markets and polls could also become a card to play in any post-election dispute. Given the conflict that followed the 2020 election between Biden and Trump — which culminated in the storming of the Capitol on 6 January — there is already fertile ground among some for the belief that the election will not go fairly. In that context, being able to point to how different the result was from what the market predicted may be all that is needed to establish scepticism and challenge the outcome.
A single bet of 150 million kronor sounds enormous in a Swedish context. But in total around 20 billion kronor is at stake on Polymarket’s exchange alone. The question is how useful markets like Kalshi and Polymarket actually are at predicting the outcome of the American presidential election. The idea of listening to large groups to forecast the future was popularised in 2004 when journalist James Surowiecki published the much-discussed book The Wisdom of Crowds. It gives examples of how well-balanced groups can reach better and faster decisions than individual experts. But not all groups display wisdom. A stock market bubble is the opposite — everyone follows each other’s behaviour, fairly uncritically, and the majority ends up losing.
Looking at how Kalshi actually works, it is quickly clear that it is not designed to aggregate balanced voices that could outperform an expert. It looks like a betting site. Choose a question — which party wins the House majority? — and based on your answer you immediately see how much money you will win if you are right. By using terms like forecasting and prediction, sites like Kalshi and Polymarket have been attributed a seriousness they may not deserve. It is a slippery slope between a political forecast with a financial payoff and an old-fashioned bet. It is wagering dressed in the trappings of political science — and wagering is, as we know, no exact science.
Taleb makes a similar observation in Fooled by Randomness: “No matter how sophisticated our choices, how good we are at dominating the odds, randomness will have the last word.” Neither polls nor prediction markets can cleanly handle the randomness and unknown factors at play in something as vast as a political election. Is it harmless to let them run? That was the American legal system’s judgement, at least. But perhaps it underestimated the reasons people bet on these questions. You don’t necessarily bet to win the wager in the moment — you bet to influence others. And then, with randomness on your side, to win in the long run.
