A Bloomberg investigation shows how cheaply a political prediction market can be moved — a single $44,000 trade sent a gubernatorial candidate's odds from the low teens to 96%. Kalshi and Polymarket have made very different bets on identity verification and traceability, and the real risk isn't that markets lie, it's that thin, early markets get treated as forecasts before anyone checks how little it costs to distort them.
On the morning of January 29, San Jose Mayor Matt Mahan officially entered California's crowded governor's race. He had Silicon Valley backing and a working-class origin story, but little name recognition next to better-funded rivals. That same morning, a Polymarket account called Fine-Tributary spent $44,000 buying "yes" contracts on Mahan winning — a bet large enough, relative to how thin the market still was, to push his implied odds from the low teens to 96 cents on the dollar. Traditional polling had him closer to 4%.
Other traders noticed the gap and sold back in, dragging the price down. But a day later, Polymarket still showed Mahan with a 36% chance of winning the race — nearly nine times what the polls said. This is the case study at the center of a Bloomberg Businessweek investigation published July 29, and it's a remarkably clean demonstration of a problem prediction markets have mostly gotten to wave away: it doesn't take much money to move one.
Prediction markets have spent the past two years building a reputation as a "truth serum" for forecasting — a claim that rests entirely on the idea that real money disciplines the price, because people who are wrong lose it. Media outlets increasingly treat market odds as a data point alongside polling, sometimes citing markets with startlingly thin volume. Bloomberg's Big Take podcast noted that a market for the next Speaker of the House had drawn just $16,000 in total trading across its entire lifetime before outlets started citing its odds as meaningful.
That thinness is exactly the vulnerability. A market with a few hundred dollars a day in ordinary trading doesn't require a well-funded conspiracy to distort — it requires a spare $44,000, which is a rounding error against most campaign budgets and well within reach of any moderately wealthy individual with an opinion. The Mahan trade wasn't even the most extreme example Bloomberg found: the newsroom's review of tens of thousands of flagged transactions across Kalshi and Polymarket surfaced a broader pattern of trades bearing the hallmarks of manipulation or improperly sourced information becoming more common on Polymarket starting this January.
This is happening against a backdrop of extraordinary growth. Combined monthly trading volume across Kalshi and Polymarket rose from under $5 billion in September 2025 to roughly $24 billion by April 2026, according to Pew Research Center's analysis of data from The Block — a pace of growth that, by comparison, now exceeds the roughly $14 billion Americans wager monthly through legal U.S. sportsbooks. The category isn't a curiosity anymore. It's mainstream financial infrastructure that political operatives, journalists, and increasingly campaign donors treat as a live signal.
Kalshi and Polymarket have taken structurally different approaches to the same underlying risk, and the difference matters more than most coverage credits it. Kalshi is a CFTC-regulated Designated Contract Market that performs identity verification on its users. Polymarket's international platform — the larger of its two exchanges and the one Bloomberg's investigation focused on — has historically operated with far less identity verification, settling trades in a stablecoin on a public blockchain where transactions are visible but the humans behind wallet addresses often aren't.
That transparency cuts both ways. Bloomberg's analytics partner, Polysights, was able to flag roughly 34,000 Polymarket transactions between August 2025 and June 2026 that fit patterns associated with insider trading or manipulation — newly created accounts, unusually large bets relative to typical market size, entries at unusually favorable odds. That kind of pattern analysis is only possible because Polymarket's global exchange is public by design. Kalshi's identity-verified, dollar-settled structure makes anonymous manipulation harder to execute in the first place, but it also means outside researchers can't independently audit its trade flow the way they can Polymarket's. Both platforms have recently tightened their rules: Kalshi says it screens to block politicians from betting on their own races and now collects employment data in some markets to guard against insider trading, while Polymarket updated its rules in late March to prohibit trades based on stolen confidential information or wagers by anyone positioned to influence an outcome.
The strongest case for prediction markets is baked into the Mahan example itself: the market didn't stay at 96 cents. Other traders saw the mispricing and bought back in within about a day, pulling the price down substantially. Academic researchers — notably economists Robin Hanson and Ryan Oprea, whose work on market manipulation is frequently cited by prediction-market defenders — have argued that manipulation attempts can, somewhat counterintuitively, improve long-run price accuracy, because they invite exactly the kind of corrective trading that punishes the manipulator and pulls the price back toward the truth.
There's real evidence for this beyond theory. A separate, previously documented case involved a trader spending over $11,000 to accumulate roughly 140,000 contracts at eight cents each, chasing a potential $130,000 payout on a mispriced outcome — an attempt that ultimately failed and cost the trader several thousand dollars once other participants corrected the price. Manipulation, in other words, is not free money; it's a bet against a market that tends to notice.
But "the market eventually corrected" is a weaker defense than it sounds, for two reasons the self-correction argument tends to skip past. First, correction took roughly a day — plenty of time for a screenshot of Mahan at 96% to circulate, get cited by a campaign or a cable news chyron, and shape a news cycle before the price ever moved back. Second, even after correction, the market settled at 36%, not anywhere near the 4% polling baseline — a gap that never closed on the timeline Bloomberg examined. Resilience over a period of days is real. It is not the same claim as accuracy at the moment a reader or a headline actually looks at the number.
The honest synthesis here cuts against both the boosters and the skeptics. Prediction markets are not lying about what they measure: a price genuinely reflects what traders are currently willing to pay, and that number is real, auditable, and hard to fake at scale over time. The mistake — made constantly by the media outlets now treating markets as a polling substitute — is quietly swapping "what traders are willing to pay right now" for "what will actually happen," as though those were the same claim. They converge over long, liquid, well-traded markets. They diverge sharply in exactly the conditions that make for a viral chart: a fresh, thin, early-stage market on a lightly polled long-shot candidate, which is precisely what Mahan's gubernatorial bid was in late January.
This is a live regulatory question, not just an academic one. The CFTC has separately been warning platforms about the risks of blanket self-certification for new event contracts, a thread that runs parallel to the manipulation question — both are, at bottom, about whether the market's plumbing can be trusted at the speed the public wants to use its output. As prediction markets scale further into mainstream financial products, including proposed equity-index contracts and corporate hedging instruments, the amount of money required to distort a thin early market becomes a bigger liability, not a smaller one — because more people will be watching the number before the correction has time to happen.
Mahan's gubernatorial run didn't ultimately go anywhere; better-funded, better-known candidates in the field eventually crowded him out, and his Polymarket bump was, in the end, a one-day story rather than a campaign-defining one. But the mechanism that produced it — a legally regulated market, moved 82 points by an amount of money that would barely cover a month of digital ads — is still sitting underneath every thinly traded contract prediction markets host today. The question worth asking about any market-derived headline isn't whether the price is fake. It's whether anyone checked how little it would have cost to make it say something else.
Grad PLUS lent up to the full cost of attendance. Its replacement caps professional students at $50,000 a year — against a medical-school year that averages $61,110. The difference does not disappear.
Fitch tracks sixteen perpetual non-traded BDCs. In Q2 2026, redemption requests averaged 10.3% of shares against a standard 5% cap, and ten of the sixteen breached it. Nothing broke — which is the uncomfortable part.
Nationally, asking prices are down about 2% and inventory is up 3.2% — a picture of orderly rebalancing. Underneath it, Austin's median list price is down 12.2% while San Francisco's inventory has fallen nearly 20%. Those are not variations around a mean.