When Predictions Become News
The CNN-Kalshi Deal and What Might Happen When Elections and Markets Collide
It’s October, 2028. JD Vance and Mark Cuban are locked in a virtual tie for the presidential election. Suddenly, Vance’s price starts surging on prediction markets. CNN, which offers breathless, round-the-clock coverage of prediction-market prices through its partnership with Kalshi, documents the spike in great detail.
Meanwhile, no one knows why the price started spiking in the first place. Democrats insist the market is “rigged.” They point to evidence that a large batch of suspicious trades moved the market without any new polls or other apparent new reason to start favoring Vance.
The New York Times runs a story claiming that traders backed by the Saudi Arabian sovereign wealth fund are placing large bets on the election markets in order to drive favorable Vance coverage on CNN. Republicans say the price is justified, point out that there’s no evidence price spikes influence election outcome anyways, and accuse Democrats of trying to stifle free speech and censor real information about the election. No one can immediately figure out what’s true.
I’ve studied prediction markets and elections for a while. In this piece, I want to explain why a scenario like the above is plausible—maybe even likely—in the coming years, even though successful manipulations of prediction markets are rare and evidence that they affect voter behavior is scant.
As I’ll explain, attempts to manipulate these markets are inevitable. And when they happen, the political fallout may vastly outweigh the direct effect on election outcomes. In a polarized environment primed to see every anomaly as a conspiracy, even a short‑lived distortion can ignite accusations of foreign interference, corruption, or elite collusion. If 2016’s “Russian ads on Facebook” scandal taught us anything, it’s that the secondary effects—the panic, the accusations, the erosion of trust—can eclipse the actual impact of the underlying act.
Yet abandoning prediction markets would be the wrong lesson. As traditional polling becomes more brittle in an AI‑saturated environment—where response rates are extraordinarily low, and pollsters are struggling to detect AI responses amongst real human respondents—prediction markets offer a useful supplementary signal that integrates dispersed information with real financial incentives.
The challenge, to expand on my previous piece on prediction markets, is governance: building systems that preserve the informational value of prediction markets while reducing opportunities for abuse. That might mean making sure broadcasters focus on reporting on the thicker markets that are harder to manipulate, encouraging platforms to monitor for signs of coordinated manipulation, and a cultural shift toward interpreting market moves with humility rather than alarmism. If we get that right, prediction markets can evolve into a sturdier, more transparent component of our political information ecosystem—one that helps us understand elections without becoming a new vector for distrust.
History tell us: expect people to try to manipulate the markets
“All eyes are fixed on the betting market now. Its fluctuations are followed with feverish interest by the great crowd of ordinary voters, who have no way personally of finding out the drift of public sentiment, and depend slavishly on the opinions of men who make a business of wagering hundreds of thousands of dollars on every election.”
–Washington Post, Sunday November 5th, 1905
In the 1916 presidential election, Charles Evans Hughes led Woodrow Wilson in the New York betting market. Remarkably,the news regularly reported on the betting markets in this era of American politics. And because they were covered in the news, the specter of manipulation loomed large. In 1916, not wanting to be perceived as behind, Democrats claimed the market was “rigged,” which the press duly reported on (see below).
The possible threat of manipulation has never gone away. On the morning of October 23rd, 2012, during the campaign between Barack Obama and Mitt Romney, a trader placed a large order for Romney shares on InTrade, causing his price to spike about 8 points, from a bit below 41 cents to nearly 49 cents—indicating a nearly tied race, if you believed the price. But the price quickly reverted, the media barely noticed, and the world kept on turning. The identity of the would-be manipulator or manipulators has never been confirmed, to my knowledge.
Sometimes, though, you can even see people spelling out the logic for their attempted manipulations openly. A fascinating study from 2004 chronicles a case of intentional market manipulation in the Berlin 1999 state elections. The authors quote a real email sent by the local party headquarters urging its members to bet in the prediction market:
“The Tagesspiegel [one of Germany’s largest newspapers] is publishing a PSM [Political Stock Market] on a daily basis, according to which the FDP [Free Democratic Party] is traded at 4.23% at the moment. You find the PSM on the Internet at http:// berlin.wahlstreet.de. Many citizens do not consider the PSM as a game, but as a result of opinion polls. Hence, it is important that the price of the FDP is going to rise during the last days. As is the case with every exchange, the price level is a result of the demand. Please participate at the PSM and buy FDP contracts. Eventually, we are all convinced of the success of our party.”
These fears came up in 2024, too. In the run-up to the election, the Wall Street Journal ran a piece questioning whether Trump’s advantage on Polymarket, which seemed much larger than his advantage in the polls, could be the result of undue influence: “The big bets on Trump aren’t necessarily nefarious. Some observers have suggested that they were simply placed by a large bettor convinced that Trump will win and looking for a big payday. Others, however, see the bets as an influence campaign designed to fuel social-media buzz for the former president.”
The scrutiny in 2024 was especially telling, too, because it conjured fears of foreign influence. As it turned out, the bets driving the Polymarket price up were from a French investor—but despite speculation, there is little reason to think it was about manipulation. In fact, the investor commissioned his own private polls and seems to have been focused on making money, not manipulating the market.
Two themes emerge from this history. First, attacks are common and we should expect them to occur again in the future. And second, even if the attacks often don’t work, strategic actors can conjure up fears about them.
How much do these attacks matter?
Whether these efforts will influence voters’ behavior depends on two factors: can the manipulation meaningfully move the price in the market, and does moving the price affect voter behavior?
Let’s start by tackling the logic for why manipulating the market, if you could do it, would supposedly help you achieve your political goals—because it’s not as obvious as people think.
Steel-manning the case for feedback loops
Here are two ways prediction markets could affect the elections they’re supposed to measure.
The Bandwagon Effect
The bandwagon effect is the idea that voters gravitate toward a candidate who appears to be winning, whether because of conformity, the satisfaction of backing a winner, or the belief that market odds signal candidate quality.
If being favored helps a candidate become favored, then broadcasting prediction-market prices on the news creates an incentive to push those prices upward. A manipulator might try to inflate their preferred candidate’s odds, hoping to trigger a feedback loop: higher market price → voters perceive momentum → voters shift support → price rises again. The market becomes a self-fulfilling prophecy.
This is often the core worry behind manipulation fears. In my Vance-Cuban example, the manipulator’s bet is that making Vance look stronger will help him become stronger.
The Complacency Effect
On the other hand, voters might stay home when their candidate looks safely ahead. If the race seems close, or their candidate appears to be losing, they could be more motivated to turn out. In that world, widely broadcast prediction-market prices create a market pressure keeping the odds near 50-50. If the market starts to favor one candidate, traders know that that candidate’s supporters start to lose enthusiasm, pushing the price back down.
This, too, opens the door to manipulation. A frontrunner worried about complacent supporters could quietly buy shares of her opponent to tighten the market and signal a closer race. Conversely, supporters of the trailing candidate might push her price even lower to lull the other side into staying home. In this case, the market becomes a self-refuting prophecy: the signal designed to reflect expectations instead works to overturn them.
Although it’s very much disputed, some people claim that Brexit is an example of this phenomenon. As one LSE report put it, “It is well known that polls affect both turnout and voting, particularly when it looks as though a particular result is a foregone conclusion. It seems…[that] more Remain than Leave supporters…chose the easier option not to vote, probably because they believed that Remain would win.”
Evidence suggests voters don’t care much about how close the election is
But here’s the thing. If bandwagon or complacency effects exist, the evidence suggests they’re typically small. American elections are fairly stable—driven mostly by partisanship and fundamentals like the economy—so if voters were highly reactive to narratives about who’s in the lead, outcomes would look far more chaotic. And when researchers have tried to directly shift people’s beliefs about closeness or pivotality, the behavioral effects are consistently modest.
Take the cleanest test we have of the theory that voters will turn out more if the election is closer: Enos and Fowler’s study of a Massachusetts state legislative race that literally ended in a tie. In the required re-run, they randomly told some voters that their district’s previous election had come down to a single vote. Even this extreme treatment barely moved turnout.
Likewise, Gerber et al showed voters different poll results in large field experiments. People updated their beliefs about how close the election was, but turnout barely shifted. A study of Swiss referendums found a slightly bigger but still very modest effect: in that context, heavily publicized close polls seem to bump turnout, but only by a few percentage points.
It’s possible that signals of electoral closeness do move some voters some of the time, but the effects are probably measured more in inches than in miles. This doesn’t mean we shouldn’t worry about manipulation, but it does mean we should be looking for small effects in hyper-close elections, not distortions that turn a close election into a landslide.
Manipulating thick markets is hard and expensive
This brings us to the second question: how hard is it to actually manipulate prediction-market prices? The answer, as the Romney example suggests, is pretty hard.
Rhode and Strumpf’s study of the Iowa Electronic Markets during the 2000 election found that manipulation attempts were costly and short-lived. In one clear case, a trader repeatedly blasted the market with oversized buy orders to push prices toward a preferred candidate. Each push briefly moved the odds—only to be immediately erased as other traders arbitraged the distortion and snapped prices back to fundamentals. The manipulator spent heavily and lost money, while the market showed strong mean reversion and resilience.
In the Vance–Cuban hypothetical, this matters a lot. Moving a presidential market in October would require serious cash, and plenty of traders would be waiting to trade the spike back down. The blip might last long enough to flash on CNN, but by the time Anderson Cooper is talking about it, the price may already be back where it started.
But when markets get thin, the story changes. In low-liquidity settings, researchers have shown it’s possible to budge long-term prices—not because traders are convinced, but because no one’s around to push back.
Some ideas on what to do
Maybe evidence suggests that manipulation in major election markets is unlikely to have big effects, but that doesn’t mean we should stand pat. In this new world where prediction markets are fusing with social media and cable news, it’s possible that manipulating prices could matter more than in the past. And even if they don’t, the fears that they could threaten to cloud our shared perceptions of the integrity of our political system. So what should we do to get in front of this issue?
For Broadcasters
Here are a few basic ideas for how broadcasters can help maintain the integrity of the information environment around elections as they report on prediction markets.
Implement liquidity floors. CNN and other news stations that report on prediction market prices around elections and other political events should focus on thick markets where the price is likely to reflect more accurate expectations and the cost of strategic manipulation is high. They should not report on prices from illiquid markets where the prices are less likely to be accurate and the cost of manipulation is low.
Here are two examples to make this idea clear.
The first image shows the resting order book on Kalshi for the contract predicting JD Vance as the winner of the next presidential election (as of the afternoon of Dec 9th). You can see that the bid/ask spread—the gap between the bids people have submitted to sell yes on Vance and the asks they’ve submitted to buy Vance—is tight. Next to each offer shows the number of shares available at that price. To momentarily move the last realized price from 30 cents to 33 cents, a trader would have to spend a little more than 30,000 dollars to sweep the asks from 30 through 33 cents (and this would presumably be instantly traded back down). This is a reasonably thick market, and given the tight bid/ask spread, reporting on it while flagging the uncertainty seems reasonable to me.
The second image shows the resting order book for the 2028 Ohio Senate race—perhaps a newsworthy event, but not one with meaningful liquidity on Kalshi yet. The bid/ask spread is much wider, the number of shares available is much smaller, and it would be quite cheap to move the last realized price by a large amount. This is not a market I would want to report on with any confidence.
Incorporate other signals of electoral expectations. In addition to focusing on these more liquid markets, news stations should keep an eye on polls and other indicators of electoral expectations that, while they have other drawbacks, are less prone to strategic manipulation. If they see a major deviation between market prices and other signals, they should look for evidence of manipulation.
For Prediction Markets
Platforms can also do a lot to help keep broadcasters and the public informed about the quality and integrity of the prices they’re seeing.
Build monitoring capacity. Systems and staff who can detect spoofing, wash-like trading, sudden one-sided bursts in thin books, and coordinated account activity. Companies like Kalshi and Polymarket may already have some of these capabilities, but likely can invest more resources in these areas, following the growth of similar teams at AI and social media companies, if they want to be seen as responsible players in the information ecosystem.
Consider stepping in when there are violent, unexplained price moves. This includes simple circuit breakers for abrupt moves in thin markets, as well as integrity halts followed by call auctions to reset prices when something looks artificial.
Report pricing metrics more resistant to manipulation. Use time-weighted or volume-weighted prices for anything that’s going to be displayed on TV, rather than raw last trades.
Continue providing increased transparency around trades. Just as important is transparency: publishing measures of liquidity, concentration, and anomalous trading patterns—without identifying individuals—so that journalists and the public can understand when a price move reflects real information versus shallow order-book noise. Major markets like Kalshi and Polymarket already display the order book, but more detailed metrics and dashboards that are digestible to the public would be useful.
For Policymakers
Policymakers should adapt tools already used in financial markets to this new class of politically sensitive event contracts.
Police market manipulation. The first step is clarity: make explicit that intentional attempts to distort the prices of election prediction markets—as a way to sway public opinion or media coverage—fall under existing anti-manipulation authorities. Regulators can then prioritize rapid coordination between CFTC, DOJ, and election-security agencies when large, unexplained price moves occur in the run-up to an election.
Regulate foreign and domestic political involvement in markets. Finally, because election markets are uniquely exposed to foreign influence and campaign-finance concerns, policymakers should consider two safeguards: (1) monitoring for foreign manipulation by tracking the nationality of traders, which should already be doable thanks to KYC laws that predication markets operate under in the US; and (2) disclosure rules or prohibitions for campaigns, PACs, and senior political staff. If spending money to manipulate a price is functionally an unreported political expenditure, regulators should treat it as such.
Conclusion
Prediction markets can help make our elections clearer rather than more confusing—but only if we set them up responsibly. The CNN–Kalshi partnership points to a future where market signals sit alongside polls, models, and reporting as part of our shared political information environment. That’s a real opportunity: in an AI-saturated world, we need tools that can surface dispersed knowledge without adding distortion. But that promise depends on good governance, including liquidity standards, monitoring, transparency, and a more measured way of interpreting market movements. If we get those pieces right, prediction markets can improve how we understand elections and support a healthier democratic ecosystem in an algorithmic age.
Disclosures: I receive consulting income from a16z crypto, a venture capital company that recently announced an investment in Kalshi. I also receive consulting income from Meta Platforms, Inc.