When Donald Trump won the presidential election in 2016, the polling institutes were left with eggs in their faces. Nate Silver’s FiveThirtyEight gave Hillary Clinton a 71.4% chance of victory. The Princeton Election Consortium went even further – a whopping 99%. But at Betfair, the world’s largest betting exchange, Trump’s odds were ticking upwards in the weeks before the election. The odds reflected something the polls did not capture. And this is really no wonder, because just as experienced players at the best online casinos know that the house always has an edge, so the betting markets know that money on the table forces people to think more sharply than any survey can.
Money talks – polls babble
There’s an old saying in the gambling world: “Put your money where your mouth is.” And that’s really the whole point here. When you respond to an Ipsos or Gallup poll, it costs you nothing to say what you believe. You can lie, exaggerate, or just answer what you think sounds good. No consequences.
On Polymarket, on the other hand, people trade contracts with real dollars. During the 2022 US midterm elections, traders put over $300 million in political contracts. It’s not pocket junk. When you risk your savings on Raphael Warnock winning in Georgia, you google a little extra, check electoral districts, and consider early voting numbers. You do your research because your wallet depends on it.
Researchers at Columbia University published a study in 2023 that compared the predictions of betting markets with aggregated polls over five election periods. The result was quite clear:
- Betting markets were the winner in 74% of close state elections, defined as margins below 5 percentage points. The opinion polls landed at 67%. That may not sound like a huge difference, but in an election where one state can tilt the entire result – think Pennsylvania 2020 – 7 percentage points is worth its weight in gold.
- Markets adjusted faster after news events. When the FBI reopened its investigation into Clinton’s emails in October 2016, Betfair reacted within hours. The opinion polls took three to five days to capture the effect.
- Betting markets were less prone to hardening, i.e. the tendency for pollsters to copy each other’s methods and end up with almost identical numbers that all miss in the same way.
Wisdom of crowds – but only when it costs something
The idea behind “wisdom of crowds” is not new. James Surowiecki wrote the book about it in 2004. But his point is often misunderstood. Crowds are only smart under specific conditions – people need to have independent opinions, there needs to be diversity of perspectives, and there needs to be a mechanism that aggregates information effectively.
Betting markets meet all three requirements. Polymarket has traders from over 100 countries. Some are political nerds, others are quantitative analysts who run their own models, and some are just people with a good gut feeling and some cash to spare. The price of a contract—let’s say “Will Biden win Michigan?” at 62 cents—represents the collective wisdom of thousands of people who all have something to lose.
Opinion polls do not have this mechanism. They ask a random sample of people and hope that the sample is representative. The problem is that it rarely is, and we know that.
What Political Science Steals from Sports Betting
What is fascinating is that political scientists are now actively borrowing methodology from the gambling industry. Nate Silver, who started his career building the PECOTA model for baseball statistics, has always been open about this link. His selection model uses the principles of Elo rating, which was originally developed for chess and later adopted by sports betting companies.
But it doesn’t stop there. Here are three concrete methods that have migrated from sports betting to election forecasting:
- Real-time Bayesian update. For years, sportsbooks have been continuously adjusting odds based on new information – an injury, a red card, a change in weather. Choice models are now doing the same. When a new measurement from Marist or Quinnipiac comes in, the models update their probabilities immediately, weighted by the historical accuracy of the measurement.
- Sharpness weighting. In sports betting, bookies distinguish between “sharp money” and “square money.” Sharp bettors have a track record of profit, and when they move, the lines follow. PredictIt began to implement similar logic – traders with historical accuracy implicitly gained more weight in the market price because they traded larger volumes.
- Correlation modelling between markets. Just as a sportsbook knows that an NBA semifinal favorite also influences the championship futures market, election forecasters now model how the outcome in one state correlates to others. If a Republican surprises in Pennsylvania, what does it mean for Wisconsin and Michigan? This type of correlation analysis came straight from the sports betting playbook.
2024 – When Polymarket crushed everyone
During the 2024 presidential election, Polymarket became the unofficial temperature gauge for the entire election. In October, the platform had over $3.5 billion in trading volume on the presidential election alone. Trump contracts were trading at around 60 cents while most national polls showed a steady run. Polymarket was right. Not only about the winner, but the market priced in a wider margin than what the measurements suggested.
That doesn’t mean that betting markets are infallible. They’re better – but they’re not perfect.
- Liquidity issues can distort prices. On PredictIt, where there was a limit of $850 per contract, small groups of ideologically motivated traders could push prices in a direction that didn’t reflect reality.
- Manipulation attempts occur. In 2024, the Wall Street Journal reported on a single trader who placed over $30 million in Trump contracts on Polymarket. It sparked debate about whether the price reflected genuine information or just one rich person’s beliefs.
Betting markets have their weaknesses, but people risking their own money consistently beat those who only answer a phone. Skin in the game changes everything – Nassim Taleb knew that a long time ago. Opinion polls are not disappearing, but they are becoming one of many sources of input. The smartest forecasters are already using metrics such as raw data and betting markets as calibration. It’s the future of election forecasting.
