California’s races show prediction markets aren’t magic

California’s prediction – Prediction markets got some California numbers right—like who led in two major races—but missed key outcomes about who would ultimately qualify for the November general election, raising questions about how campaigns and voters should treat these wagers.

In the weeks leading up to California’s gubernatorial and Los Angeles mayoral contests, millions of dollars of belief were poured into prediction markets—bets that claim to turn uncertainty into something measurable.

On paper, it looked simple: if traders collectively thought a candidate would finish first, the market would show it. If they didn’t, the market would flinch.

In practice, the markets behaved less like a crystal ball and more like a coin with a memory—sometimes sharp, sometimes off by just enough to matter.

Former Health and Human Services Secretary Xavier Becerra was widely expected to take the top spot in California’s governor’s race. Incumbent Los Angeles Mayor Karen Bass was also a leading pick to win her reelection bid. In those “who finishes first” predictions, the markets were fairly accurate.

But when the question shifted from who would lead to who would join Becerra and Bass in the November general election, the confidence thinned.

In the governor’s race, billionaire Tom Steyer was second on the prediction markets. Yet Steyer finished third—behind Becerra and the Republican, former Fox News host Steve Hilton—despite spending more than $200 million of his own money.

Betters on Kalshi mostly nailed the top three in that race. Still, the overall outcome underscored a hard reality for anyone watching these markets closely: being close on the board is not the same as being right when voters and candidates ultimately decide.

The Los Angeles mayoral race followed its own pattern. Prediction markets typically saw Republican and reality TV star Spencer Pratt running behind Bass. Ultimately, it was Los Angeles City Councilwoman Nithya Raman who took the spot to face Bass in the general election.

The discrepancy has a temptation built in—people see “prediction markets” and assume they should behave like polls. They don’t.

“These can make a relevant claim about their accuracy,” Robin Hanson, an associate professor of economics at George Mason University, said. “What we see is that unmodified markets are better than unmodified polls.”

That distinction matters because polling and markets are built for different kinds of answers. A poll measures what a sample of voters says they plan to do. Prediction markets measure what traders collectively believe will actually happen on Election Day.

Hanson put it plainly: “Polls versus markets isn’t really the best framing. A poll is an input into an estimation process, and the market is a forum in which different people using different estimation processes compete.”

There’s also a basic demographic mismatch. Polling samples potential voters from specific areas. Prediction markets are open to anyone in the world, meaning participants don’t have to live in the state, don’t have to be eligible to vote, and don’t even have to be 18 years of age and up.

Darry Sragow, a longtime California political consultant and Democratic strategist, said that can distort how people interpret the signal. “You have people playing prediction markets who don’t have to live in the United States. they certainly don’t have to live in California. they don’t have to be 18 years of age and up. meaning they’re not necessarily eligible to vote. ” he said.

It’s part of why the comparison to campaign polling can be misleading. Prediction markets didn’t arrive fully formed in American politics overnight.

Kalshi wasn’t legalized in the U.S. until late 2020, while Polymarket didn’t get legal status until late 2025. That is newer than the long arc of modern presidential polling. which began roughly 90 years ago when George Gallup’s polling method correctly predicted the election of former President Franklin D. Roosevelt.

Sragow drew another sharp line between what markets measure and what campaigns need. “Private polling and opinion research, including focus groups, has nothing to do with predicting who’s going to win. It’s about crafting a winning campaign.”

So should campaigns now treat prediction markets the way they treat polls—an instrument for strategy and message? Sragow’s answer was cautious. “It’s too early to tell,” he said. “It’s a new phenomenon.”

Even if the markets are imperfect predictors, there’s another question lurking behind the numbers: what they might change.

Prediction markets tend to list who the general public believes will win, and voters who dislike uncertainty often gravitate toward momentum. Sragow described a belief that tends to show up in campaigns but rarely gets discussed out loud.

“There is this underlying phenomenon that’s very evident when you’re running a campaign, but doesn’t get talked about a lot publicly, and that is that the voters want to pick a winner,” he said. “They hate picking a loser.”

California’s results didn’t settle that question. They sharpen it.

The markets were strong when the race was about who would take the first spot. They were weaker when the race turned into a test of who would move through to November. And for anyone placing political bets—whether they call them analysts. traders. or voters—those two outcomes may be the clearest lesson the season has delivered so far: prediction markets can forecast parts of an election. but they can’t replace the messy decisions that actually determine it.

prediction markets Kalshi Polymarket California governor's race Los Angeles mayoral race Xavier Becerra Karen Bass Tom Steyer Steve Hilton Nithya Raman Spencer Pratt George Mason University Robin Hanson Tarek Mansour Darry Sragow November general election