Political and business media, pundits, and experts continue to cite prediction market odds as a credible forecast for the upcoming presidential election, regularly pointing to betting odds on prediction markets such as Polymarket, Kalshi, PredictIt, and Interactive Brokers as a guide to who is up and down in the horse race – but that is a big mistake.
It appears few journalists, pundits and experts have dug deep into how these prediction markets actually work – and fewer yet have ever actively ‘bet’ on these sites. We have, and we’ve been startled by what we found as we made bets on these sites (as an academic exercise).
Here are five key reasons why these prediction markets should not be cited by media as credible, reliable indicators.
1. Political prediction markets have extremely thin volume and liquidity – meaning very few traders actually bet on them
It seems that there might have been more stories written about the prediction markets than people who actually use them – as we were shocked by how few traders actually bet on these platforms, and the thin trading volume and liquidity which make it extremely difficult to make bets to begin with.
On some platforms, we tried to make bets on several key battleground races – only to find out we couldn’t, because there were zero sellers in those markets. In other markets, the spread – meaning the difference between the bid price and the ask price – was so massive, running upwards of 50%, that it was substantive impossible to buy or sell. (some screenshots of examples are linked). All this means that the so-called price cited by media accounts is merely a phantom figure and not representative of reality, since nobody can buy or sell in these markets in practice. On other platforms, even the largest market– the presidential election odds – feature extremely low volume, generally in the tens of thousands of dollars a day but sometimes hovering in the lower hundreds of thousand of dollars a day, to the point where a single small bet of a few thousand dollars can immediately move the market by several percentage points in real time. It can be substantively impossible for anyone to bet more than a few thousand dollars without immediately moving the markets – which might explain why the betting odds are often so disparate across different prediction market sites, and don’t even move in alignment with each other much less with polling results.
2. Political prediction markets are rife with unchecked foreign manipulation – and it is too easy to manipulate these markets.
The only prediction market with significant scale – Polymarket – is open only to foreign individuals, as US citizens are banned from trading there. But even though Polymarket has deeper liquidity than other prediction market sites, it is still extremely easy for deep-pocketed foreigners to single-handedly move markets. The Wall Street Journal reported that the rise in Trump’s odds on Polymarket over the last few weeks was almost single-handedly fueled by a series of $30 million bets, apparently from the same foreign entity. One shudders to think which foreign entity might be behind this naked market manipulation.
The lack of liquidity and lack of trading volume makes it all too easy to manipulate these prediction markets for anyone who wants to, foreign or domestic. For example, the total historical trading volume for the critical Montana Senate Race – which control of the Senate depends on –amounts to only ~$75,000 on Kalshi, similar to peer platforms. This means for only a few tens of thousands of dollars – pocket change for a wealthy donor or a foreign oligarch – one could easily corner this market by either buying or dumping shares, creating momentum either for or against their chosen candidate.
2. Political prediction markets are rife with unchecked foreign manipulation – and it is too easy to manipulate these markets.
The only prediction market with significant scale – Polymarket – is open only to foreign individuals, as US citizens are banned from trading there. But even though Polymarket has deeper liquidity than other prediction market sites, it is still extremely easy for deep-pocketed foreigners to single-handedly move markets. The Wall Street Journal reported that the rise in Trump’s odds on Polymarket over the last few weeks was almost single-handedly fueled by a series of $30 million bets, apparently from the same foreign entity. One shudders to think which foreign entity might be behind this naked market manipulation.
The lack of liquidity and lack of trading volume makes it all too easy to manipulate these prediction markets for anyone who wants to, foreign or domestic. For example, the total historical trading volume for the critical Montana Senate Race – which control of the Senate depends on –amounts to only ~$75,000 on Kalshi, similar to peer platforms. This means for only a few tens of thousands of dollars – pocket change for a wealthy donor or a foreign oligarch – one could easily corner this market by either buying or dumping shares, creating momentum either for or against their chosen candidate.
3. Extremely dubious governance practices – raising red flags about credibility and reliability
The governance of these prediction market platforms raises serious red flags with abundant personal conflicts of interest – apparently blurring the lines between personal political agendas and business judgment. For example, Polymarket’s CEO Shayne Coplan is notably close to the Trump camp and a dinner partner of Donald Trump Jr. and other top Trump deputies, and bragged about schmoozing Trump allies at the Republican National Convention. Furthermore, the largest investors in prediction market platforms are a who’s who of top GOP donors with close links to Trump, such as Peter Thiel and Henry Kravis.
There are also red flags about whether the practices of these platforms align with the best interests of their own users. For example, PredictIt charges fees on everything ranging from inactive accounts, to withdrawing money (5%), to commissions on winnings (10%); and Kalshi declined to pay customers interest on their cash positions until reversing course earlier this month. Similarly, despite advertising that Kalshi is “able to work with well-known market makers such as Susquehanna Group”, Kalshi admitted to CNBC last week that no such trades involving Susquehanna have yet occurred and that very few sophisticated institutional investors and hedge funds are using their platform.
4. Political prediction markets are of questionable legality for US citizens, and there are significant obstacles to access.
For a market to be seen as accurate and reliable, participants should be able to access those markets freely, legally, and without legal risk. That is not the case when it comes to prediction markets.
While some sites have been legalized by a recent court decision, it is technically illegal for US citizens to trade on Polymarket, the largest prediction market – as US citizens are blocked from accessing Polymarket, the result of a 2022 settlement with CFTC regulators where Polymarket was fined $1.4 million.
While some enterprising US citizens have found illicit ways around this ban, using VPNs and offshore bank accounts in tax havens to register and place bets on Polymarket, an illegal market should not be touted as representative by mainstream media and credible experts. Furthermore, it is impossible to fund a Polymarket account without owning cryptocurrency, which further restricts access to a subsection of the population.
Though technically legal, US-based platforms such as Kalshi, PredictIt, and Interactive Brokers don’t offer much greater legal clarity for US citizens, as they too operate in an ambiguous state of de facto legal limbo, wrapped up in ongoing years-long litigation against their regulator, the CFTC, which is actively seeking to shut them down. Although these platforms tout a recent favorable ruling from a federal district court judge allowing these platforms to continue to operate in the interim, the CFTC is appealing that decision, and the reality is that the Circuit Court could well rule that these platforms are illegal and shut them down in merely a few weeks’ time. This does not even begin to account for legislative pushes, led by Sen. Richard Blumenthal among others, to ban these platforms.
Few rational investors relish the prospect of getting caught up in ongoing litigation with their accounts potentially shut down and their money potentially locked up as these legal challenges play out.
5. Demonstrated lack of historical accuracy – and lack of historical viability
When it comes to accuracy, these prediction markets have an even poorer historical track record than political polling– not to mention these companies come and go with startling transience. While the current market leaders – Polymarket, Kalshi, and Interactive Brokers – had not yet been started or were inactive before this election cycle, we can learn from defunct prediction markets for a historical read.
While there were disparate results from different platforms, generally, the prediction market consensus has been wrong each of the last 3 presidential election cycles. In 2012, InTrade, the then-market leader, priced Mitt Romney to win a razor-thin victory; in 2016, PredictIt overwhelmingly priced in a Hillary Clinton victory, and similarly in 2020, Betfair priced in a Trump victory over Biden with late surges in popularity.
Despite these red flags, prediction markets have largely evaded scrutiny with media accounts, experts, and pundits citing them as credible sources of information. The reality is that prediction markets are not at all predictive, despite their name, and entirely unreliable. As Mark Cuban pithily summarized, “prediction markets are not really an indication of anything”. Not only are they of questionable legality, deterring and scaring away most US investors, but they are rife with foreign manipulation – and it is far too easy to continue to manipulate these markets with their extremely thin liquidity. Whatever these betting platforms are, they should not be cited as credible indicators of the state of the presidential race.
Correction, Oct. 22: An incorrect version of this story was originally published. It has been updated throughout to reflect edits.
