Liquidity Not a Factor in Prediction Markets’ Accuracy, Say University of Iowa Researchers
University of Iowa (Credit: Tony Webster/Wikimedia Commons)
Market size and liquidity aren’t all that important to accuracy when prediction markets attempt to forecast elections for elections.
That finding was among the more surprising findings from a new study out of the University Iowa. It casts some doubt about the utility of the new wave of commercial prediction markets, which have touted themselves “the news 2.0.” Indeed, not only is the size of operations like Kalshi and Polymarket unnecessary, other features of these giants may be undermining their accuracy.
Thomas S. Gruca and Thomas A. Rietz undertook the study, titled “The Efficiency of Binary Political Prediction Markets.” As their prototype, they use Iowa Electronic Markets (IEM), the longest-running U.S. prediction market, but a small one.
IEM uses real money but is set up as a research tool, not a commercial endeavor. It has been around since at least the mid-1990s. Trades on the platform are capped at $500, a limit which Gruca and Rietz found is more important to accuracy than total volume, because it prevents individual whales from distorting the results.
Markets Without a Position Limit Are a Different Bird
That position cap is important because it’s a major point of difference from retail prediction markets. On platforms like Kalshi, Polymarket, and others, trades can pay out in the millions of dollars. A low cap would make those platforms useless for financial hedging, the other main axis along which they claim to provide societal value. The study suggests that hedging value and predictive value are, to some extent, mutually exclusive. The authors propose that platforms like Kalshi would be better described as “event-driven markets” than “prediction markets.”
When it comes to predictive power, the theory goes that these markets reflect truth via “wisdom of the crowds.” Operators often tout themselves as more accurate than news and polls. In theory, larger crowds and greater liquidity should increase accuracy. More people with more information pour their money into the system, and the markets trend towards efficiency.
In practice, that may not be the case. The Iowa study shines a light on that, as well as some other interesting mechanics of prediction markets. It also forces one to question whether the “casino-ification” of prediction markets negatively affects their utility.
Efficiency ‘Invariant’ to Market Size, Liquidity, and Contract Count
The Iowa researchers found that, contrary to the idea that bigger crowds with more money generate more wisdom, size and liquidity had no effect on the accuracy of political prediction markets. They tested multiple “size” measurements and found none of them correlated significantly with results at the 95% confidence level.
“The size of IEM political winner-takes-all markets, in terms of the total number of active traders, the length of the market, the total market volume, and the number of different contracts offered in the market, has little impact on the efficiency of the markets in forecasting winning probabilities,” they wrote.
Importantly, they did find the markets “well-calibrated.” So, the prediction markets as reflections of truth can be a real thing. They simply don’t require hordes of users pouring ever more dollars into them in order to reach efficiency.
“[Prediction markets] neither systematically over predict nor under predict election winning probabilities across a 100-day pre-election horizon,” the paper says.
Critically, the IEM markets avoid pitfalls that can plague polls and even traditional betting markets and retail prediction markets. These include partisan bias, incumbency bias, and favorite-longshot bias.
Does Time Frame Matter For Political Prediction Markets?
The study also produced a nugget regarding timelines of the IEM political markets. It found that market accuracy wasn’t linear. That is, they didn’t trend ever closer to accuracy as they drew nearer to the date of settlement.
Again, there’s some correlation here with the volume factor. In theory, liquidity would build up toward the election, and that liquidity would help increase the market’s accuracy.
In practice, the markets did become most accurate in the final 30 days before elections. Around a week out, the prices almost perfectly reflected actual probabilities.
However, the effect wasn’t linear, as the IEM markets proved more accurate over longer timescales than intermediate ones. The point of lowest accuracy came about 60 days out from the election, roughly the point at which campaigns kick into high gear. At that point, the opening salvos of the ad war and results of the first debates can create tremendous volatility.
In practice, what this means for retail traders is that the 60-day mark may offer the greatest potential to find an edge. If markets are overreacting to outside factors during that time, then opportunities to get on the right side of an incorrect price should be plentiful.
Hedging and Gambling Activity Undermine Predictive Power
Perhaps of most importance, the Iowa paper serves as a counterpoint to the idea that retail prediction markets serve a social function as sources of truth. Proponents often tout the ability of prediction markets to inform, but the research from Iowa paints a different picture.
It actually argues that the mechanics of retail event-driven markets can, in some ways, obfuscate the truth, distorting their ability to predict future events the way a smaller prediction market can.
“Such markets blend forecasting, speculation, and hedging,” they wrote.
The latter two trading motivations interfere with retail prediction markets’ ability to forecast the future.
Speculation, in this case, is code for gambling. The researchers describe users who view prediction markets as a casino as “risk-seeking.” These users, they believe, negatively influence markets by systematically backing longshots. That causes the prices on these longshots to overstate their probabilities.
This favorite-longshot bias has long been known to affect betting markets, and research has shown the phenomenon extends to prediction markets.
Market makers on commercial platforms take advantage of the favorite-longshot bias by trading against retail users hoping for big payouts on unlikely events. For example, retail traders were still buying Yes on extreme longshots Nikola Jokic and Victor Wembanyama to win NBA MVP in May, when conventional wisdom was that Shai Gilgeous-Alexander had the award locked up (he was announced as the winner on May 17).
Hedges, Fees Further Dilute Prediction Markets
On the opposite end, highly capitalized entities can use very liquid markets to hedge risk.
Commodity Futures Trading Commission (CFTC) Chair Mike Selig consistently uses the phrase “manage risk” to justify its oversight of prediction markets. However, the Iowa researchers point out that pouring money into markets as a hedge muddies their ability to forecast the future.
That makes intuitive sense because hedging money isn’t meant to predict an outcome. It’s meant to provide a safety net against that very outcome occurring.
Fees are another factor cited by the researchers as working against efficiency in retail prediction markets. They wrote that fees charged by the platforms drive a “wedge between prices and trader beliefs.” Users may not wish to pay fees on their trades, thereby preventing them from backing their opinions with money and helping the market trend to efficiency.
The $500 cap on trades, which prevents serious hedging that could move the market away from reflecting true probabilities, also serves as an obstacle for market manipulation. Without serious money at stake, there isn’t much incentive for users to manipulate outcomes. Such occurrences have warped some retail markets. Notably, a trader allegedly made off with tens of thousands of dollars after tampering with weather in gauges in France, ensuring temperature markets would settle at long prices.
Will the Future of Prediction Markets Look More Like IEM?
The University of Iowa researchers’ fascinating findings could have ramifications down the line for prediction markets.
In the near term, prediction markets seem likely to continue to flourish. The Trump administration has reiterated its commitment to prediction markets, with Donald Trump himself recently taking to social media to affirm his backing.
Longer term, questions remain. Trump has limited time remaining in office, and Congress has put increasing pressure on prediction market operators. There’s also the matter of ongoing court cases that look likely to wind up before the Supreme Court.
Even if commercial prediction markets are ultimately deemed illegal sportsbooks, the Iowa study suggests there should be a place for the more “pure” prediction market structure. Under the right conditions, they can provide a meaningful signal. But when they become gambling outlets, they skew more towards profit factories for well-capitalized market makers.
Image credit: Tony Webster/Wikimedia Commons (license)
Players trust our reporting due to our commitment to unbiased and professional evaluations of the iGaming sector. We track hundreds of platforms and industry updates daily to ensure our news feed and leaderboards reflect the most recent market shifts. With nearly two decades of experience within iGaming, our team provides a wealth of expert knowledge. This long-standing expertise enables us to deliver thorough, reliable news and guidance to our readers.
Latest News
