Prediction markets have become a cornerstone of how analysts view elections. The theory is simple: by tapping into the wisdom of a crowd with skin in the game, we can arrive at a good estimate of the chances that a party or candidate or proposition has in an election or referendum. The science of the wisdom of crowds, but predictive markets have always had clear problems. Recently, predictive markets such as Polymarket have seen Donald Trump’s odds of beating Vice President Kamala Harris shoot up to 61%. In comparison, the Economist’s model gives Trump a 54% chance, and 538’s model gives Trump a 53% chance of winning. The difference between prediction markets and polling aggregators is the difference between claiming Trump’s victory is pretty likely, and saying the election is basically a toss-up with Trump having a slight Electoral College advantage. So why are prediction markets so confident that Trump will win? Reporting by the Financial Times shows that much of Trump’s surge on Polymarket is due to just four accounts. Wisdom of the crowds this is not.
Source: Polymarket
The FT’s reporting shows that since September, these four accounts, led by Fredi9999, have placed $30 million worth of bets on Trump winning the race for the Presidency. These four accounts hold the largest positions in the market, with their trades often making up as much as 20% of the market’s volume on some days. Theoretically, predictive markets are supposed to bring together a large pool of people with different kinds of expertise, each of whom contribute by placing bets that reflect the probability they assign to an event.
Source: Polymarket
The larger and more diverse and more informed the pool, the more accurate it can be. Predictive markets run into problems when it comes to politics, because the most informed people, politicians, their staffers, and lobbyists, are barred from betting on elections. Uninformed opinion has a large role in predictive markets. Infamously, Polymarket’s competitor, PredictIt, still gave Trump a just 10% chance of winning Wisconsin weeks after he had already won it. In addition, Texas and Nevada prohibit betting on elections, further limiting the diversity, size and informedness of the pool of people betting on elections. Finally, predictive markets have just never been very popular, so the crowd has never been very big, although liquidity has increased recently.
These limitations to predictive markets are why just four accounts have been able to trigger such a large surge in Trump’s chances. FT suggests that these four accounts could even belong to one person or group, because of the “trading and deposit patterns” which “suggest coordinations between the accounts and public comments exhibit a similar tone and writing style”. This trader or traders claims to be French, an “investor and statistician” who worked in New York between 2000 and 2006 as a trader. This person or persons claims to be non-political and to sincerely believe that Trump's odds should actually be “75% Trump”. It could be that this is a genuine bet on what this person or persons believe is a likely Trump win. Or, this could be an effort to influence public opinion, by making Trump’s victory seem certain. Nate Silver in a recent post admitted that, in the Trump era, Republicans project confidence, and Democrats project anxiety. A final possibility is that this trader is trading off of inside information, such as internal polls.
So, ultimately, we shouldn’t place too much on what polymarket is saying, because, ultimately, right now, it is reflecting the opinion of a few traders, rather than of a large, diverse, and informed crowd. If you want to assess the state of the race, stick to the Economist’s model, 538, or the Silver Bulletin model, all of which shows that this race is the tightest race in a century.
