What Can Prediction Markets Tell Us About the California Governor’s Race?

As Californians follow this year’s governor’s race, they have access not only to polls but also to data from prediction markets like Kalshi and Polymarket. Major media outlets like CNN and Fox News now include the probabilities listed on Kalshi as part of their election coverage.

And earlier this year, Polymarket announced an exclusive partnership with the newsletter platform Substack, claiming that “journalism is better when it’s backed by live markets.”

As of this story’s publishing, Becerra’s odds are listed at 74% on Polymarket and 72% on Kalshi. Steyer comes in second position on both sites, just shy of 20%.

Becerra’s odds on both prediction markets have steadily increased as more traders put money down on him becoming the next governor. The more people believe something will happen, the more valuable those “shares” become — unlike traditional betting, in which oddsmakers set odds.

But just how good are prediction markets and what can they tell us that polls can’t?

While prediction markets have advertised themselves as powerful tools to predict what will happen in the future, the experts KQED spoke with recommended looking at predictions on the governor’s race with a critical eye, as these markets can be informative but remain vulnerable to inefficiencies and manipulation.

What markets know (and don’t)

Kalshi and Polymarket let people become “traders” and make predictions about future events by putting money down on a specific outcome. There are wagers for almost anything that involves some uncertainty: whether the United States and Iran will sign a new nuclear deal, which nation will win the FIFA Men’s World Cup and even whether Jesus Christ will return — before 2027.

(But just don’t call it “betting” — both sites avoid that terminology as most states prohibit online sports gambling, which makes up the bulk of the billions of dollars traded.)

Prediction markets listing Becerra’s odds at 70% does not mean he will end up with 70% of the vote on Tuesday. Rather, this number lets traders know how much each share costs and the expected payout if he’s elected governor. If you were to buy the “yes” position on Becerra becoming the next governor, you would essentially be entering into a contract that pays $1 if he wins and $0 if he doesn’t. With 70% odds, each contract for a Becerra win will cost $.70.

As more users — and their money — join Polymarket and Kalshi to trade on the outcomes of elections, both sites have sought to brand themselves as more effective alternatives to polls. A spokesperson for Polymarket told KQED that the platform “has proved to be an accurate tool for political forecasting, oftentimes even more so than traditional polling data.”

And while Kalshi did not respond to KQED’s questions, CEO Tarek Mansour said in a recent interview that people tend to be more truthful when they have money on the line. “I think there’s a little bit of elegance in the idea that markets don’t lie,” he said.

Prediction markets can aggregate a lot of different information at once, as traders are usually keeping up with the news and different polls, said Neil Malhotra, professor of political economy at the Stanford Graduate School of Business. The assumption, he said, “is that people are incentivized to make good predictions because if they don’t, they’ll lose money.”

Traders can sell and buy shares of a race at any given time before election officials announce results. So the odds for a candidate can reflect changes on the campaign trail a lot faster than what it would take pollsters to conduct a survey.

Back in March, for example, then-Congressman Eric Swalwell saw his odds on Kalshi reach 75% but then quickly sink as more women spoke up to accuse him of sexual harassment. The day before he announced his exit from the race, Swalwell’s odds were almost at zero.

But Malhotra also cautioned that it’s still unclear how much the actions of a single trader could influence the probabilities listed on the platform. A single person could pump tens of thousands of dollars into the market for a specific outcome and potentially make one candidate look like they have a much stronger chance of victory. Earlier this year, major Democratic donor Stephen Cloobeck was blocked from Kalshi after trying to place approximately $1,000 on ally and close friend Swalwell.

And if traders are looking at different polls to make their predictions, any problems in polling, Malhotra said, “might be reflected in the prediction markets as well.”

That’s why other researchers believe that prediction markets are flawed tools.

“I think that the movements that we see day to day in these markets oftentimes should be purely ignored,” said Eben Lazarus, assistant professor of finance at the Haas School of Business at UC Berkeley.

“Markets move so much in response to small trades and small pieces of news,” he said. “That adds to the uncertainty that you should attach to the numbers that you see in them.”

Let’s run that back — on the basketball court.

By studying how basketball fans bet over NBA games, Lazarus found that individuals can be heavily influenced by small, insignificant news and actually underestimate much more important pieces of information. Participants in his study would place a lot of importance on things that would happen during the first quarter and underreact during the last few minutes of the game — when the game’s actual outcome is decided.

Traders might get really excited that Stephen Curry hits a three-pointer for the Golden State Warriors early in the first quarter, but, Lazarus said, “that means very little for the eventual outcome.”

Similar behavior comes up in other types of markets that wager on predictions, he said. “People tend to treat pieces of information as sort of too similar to one another,” he said. “Then you’re going to overreact to basically useless stuff and often underreact to stuff that matters a lot.”

That means that the odds listed in prediction markets are dependent not just on who’s betting on what outcome, but how traders process the information they’re being exposed to throughout the race. A flashy debate performance or big policy announcement may not actually move voters as much as it could influence traders who may be constantly consuming a lot more election coverage.

What’s the price of knowing?

It’s important to remember that prediction markets have also been wrong many times, said Julian Vogel, assistant professor of finance at San José State University and chartered financial analyst. “This doesn’t show you what will actually happen; it shows you what most people — who invest in this platform in particular — think might happen,” he said.

These platforms may insist that they offer data that’s accurate and unbiased, but for markets to be a “truth machine”, Vogel said, “there have to be insiders who know the truth [also] trading.”

He pointed to the case of the U.S. soldier charged with using classified information about the capture of Venezuelan leader Nicolás Maduro to make more than $400,000 on Polymarket. Having someone with access to real, classified information “tipped the scales in favor of what turned out to be the correct outcome,” he said.

“If that person would not have traded, then it would have been anybody’s best guess,” he said.

But having people with access to delicate information trade — and profit — from this knowledge raises both ethical and legal questions. The Senate banned earlier this year staff and Senators from using prediction markets, including buying shares related to federal policy. And in April, Kalshi suspended the accounts of three users who were also congressional candidates and wagered on the outcome of their own elections.

As more money flows into these sites, individual states have stepped up regulation. Last month, Minnesota became the first state to specifically ban prediction markets — including wagers over the outcome of “a federal, state, or local election.” However, the Commodity Futures Trading Commission, the federal agency tasked with regulating prediction markets, almost immediately filed a lawsuit against Minnesota over the new law, claiming the legislation undermines federal authority.

As for individual voters looking at prediction markets now, they should consider if the people making wagers are actually “learning from other people’s wisdom,” Malhotra said.

But this strategy comes with its own risks, he said. The wisdom of crowds doesn’t work when the judgment of each individual is correlated, meaning that the same outside factor influences us.

“If we’re judging a candidate by how handsome they are or something like that, rather than their quality or their policy issues … then you might actually not have the wisdom of crowds because our mistakes might be correlated with each other,” Malhotra said.

Financial markets help us understand the price of almost anything — something economists call “price discovery.” How much you pay to fill up your car’s gas tank is directly correlated with the actions of investors in petroleum markets worldwide. But with prediction data now available alongside polls, voters have the power to decide how much markets will influence their individual decisions when filling out a ballot, Malhotra said.

“A big social question we have to ask: Is the price discovery about things that are important to us as a society — like who’s going to win an election — should that be put in the same category as what’s the price of an apple?” he said.

“That’s something we have to struggle with,” he said. “What are the pros and cons of these prediction markets on our democracy?”