On March 16, a federal judge in Arizona denied Kalshi’s motion for a preliminary injunction against the state. On March 17, Arizona Attorney General Kris Mayes filed 20 misdemeanor criminal charges against the company — the first criminal charges ever filed against a major prediction market. The charges allege that Kalshi is operating an illegal gambling business in Arizona and, separately, taking bets on Arizona elections, which state law prohibits outright.
Kalshi’s response: this is a federal question, not a state one. The company operates as a Designated Contract Market under the Commodity Futures Trading Commission. CFTC Chair Michael Selig — currently the sole sitting commissioner on a body that normally has five seats — agreed, posting on X within hours that the criminal prosecution was “entirely inappropriate” and that the CFTC was “evaluating its options.” A preliminary injunction hearing is scheduled for April 3.
The legal dispute — federal preemption versus state police power, futures contract versus gambling wager — is genuinely interesting and will be litigated for years. But there’s a structural question underneath it that the law doesn’t fully capture, and that’s what I want to talk about.
What Is a Prediction Market Actually Doing?
The standard defense of prediction markets as information aggregators goes like this. Many participants, each with private information and real money at stake, trade contracts whose value depends on the outcome of a future event. The market price — the odds — reflects the collective judgment of all those participants, weighted by their willingness to put money behind their beliefs. This is, in theory, an efficient mechanism for aggregating dispersed information into a single number. It’s why Kalshi’s odds on the 2026 midterms, or on who wins the Texas Senate runoff in May, are frequently cited by journalists and campaigns as meaningful signals.
That theory assumes the market is measuring something external to itself. The odds reflect what’s going to happen; the event happens or doesn’t; the contract settles. The market and the outcome are separate systems, connected only by the information participants bring to the trade.
This assumption breaks down in political markets, and it breaks down in a specific and formally interesting way. When Kalshi shows a candidate at 8% odds, it isn’t passively recording the probability that candidate will win. It’s publishing that number to donors, staffers, journalists, and the candidate herself — all of whom will respond to it. A donor who sees 8% reconsiders the next check. A staffer starts updating their résumé. A journalist files a “fading” story. The candidate begins having private conversations about the timing of a withdrawal. Each of these responses changes the actual probability the market was trying to measure — which changes the odds — which triggers more responses.
The market is not measuring the race. The market is participating in the race. And the participation is not neutral: it systematically advantages well-funded, well-known, early-momentum candidates, because those are the candidates who generate trading volume, which generates liquidity, which generates the tighter odds that attract more attention. A candidate who might have won through a slow ground-game build can be effectively eliminated by a prediction market before the ground game has time to run.1
The Classifier That Shapes What It Classifies
Readers of an earlier post will recognize this structure. When a classification system changes, the population it classifies changes too. The IRS-ICE pipeline didn’t just measure the undocumented taxpayer population — it reshaped the filing behavior of that population, which meant the system was partly measuring its own effects. A classifier with teeth isn’t passive. It enrolls people in the classification process whether they choose to participate or not, and their response to enrollment is itself data.
A prediction market is a classifier with teeth and a ticker. It publishes its classifications in real time, to the people being classified, with financial consequences attached. The feedback loop is faster and more direct than almost any other classification system in politics. And unlike most classifiers, it is specifically designed to be visible — the whole point is that the odds are public, that anyone can see them, that the market “speaks.”
This creates what we’ve called the endogenous base rates problem: the base rate of the thing you’re trying to measure shifts in response to the measurement. A candidate’s probability of winning is not independent of what the market says her probability of winning is. The market is measuring a probability that its own publication is altering. In the limit — and this is not hypothetical, it happened in several 2024 primaries — the market prediction becomes self-fulfilling not because it was accurate but because it was influential.
Here is the formal wrinkle. The setter model tells us that whoever controls what options are on the table when a choice is made shapes the outcome, regardless of voter preferences. A prediction market that effectively eliminates candidates by making them appear nonviable is doing agenda-setting, not forecasting. It’s not telling you who will win the race as currently constituted. It’s constituting the race.
The Regulator in the Market
Now add the political economy layer, because the Kalshi story has one.
Donald Trump Jr. is a “strategic adviser” to Kalshi. President Trump is reportedly planning a crypto-based prediction market on Truth Social. The sole CFTC commissioner — a Trump appointee, occupying one of five seats on a body that currently has four vacancies — intervened publicly within hours of Arizona’s criminal charges to back Kalshi and call the prosecution “entirely inappropriate.” The Biden-era CFTC had initiated a rulemaking process to prohibit election-related prediction markets; the Trump CFTC withdrew it.
The regulator is not a neutral arbiter. The regulator is a player. And the regulator’s position — that Kalshi operates under exclusive federal jurisdiction, beyond the reach of state law — is simultaneously the legal argument Kalshi needs to survive and the commercial position that benefits the administration’s allies in the prediction market industry. The CFTC is setting the agenda for what kind of market Kalshi gets to be.2
Arizona’s AG, Kris Mayes, is a Democrat. She has made the jurisdictional question a federalism question — “no company gets to decide for itself which laws to follow” — which maps cleanly onto a 2026 campaign message about federal overreach. The April 3 hearing will produce a ruling on the preliminary injunction. Whatever it says, it will be appealed. Whatever the appeals court says, the Supreme Court will eventually be invited to weigh in on whether CFTC jurisdiction preempts state gambling law.
While that litigation proceeds, Kalshi’s election markets keep running. Users are currently betting on the 2026 Arizona governor’s race — the race in which Kris Mayes’s political future is arguably implicated — and on the 2026 Arizona attorney general’s race, in which Mayes herself may be a factor. The platform accused of illegally influencing electoral outcomes is publishing odds on the elections of the people prosecuting it. The classifier is classifying its classifiers.
None of this resolves the legal question, which is genuinely hard and will take years. What it does is clarify what kind of question it is. The debate about whether Kalshi is a futures exchange or a bookmaker is a debate about where to draw a line. The more interesting question — whether a real-time, publicly-visible, financially-consequential classifier of electoral probability should be subject to any democratic accountability for its effects on the elections it classifies — isn’t really about gambling law at all.
We’ll have more to say about this as the April 3 hearing approaches. In the meantime, I leave you with this.3
1 The 2024 Biden-Harris transition offers the clearest recent example. Prediction market odds on Biden’s reelection collapsed well before his withdrawal, accelerating the donor and endorsement flight that made withdrawal feel inevitable. Whether the market caused the withdrawal or predicted it is the kind of counterfactual that’s genuinely hard to resolve — which is exactly the problem. A classifier whose effects can’t be distinguished from its measurements is not giving you information. It’s giving you a fait accompli.
2 This is also a conservation of impossibility story: the federal preemption argument moves the regulatory question from a domain where democratic accountability exists (state legislatures, state AGs, state gaming commissions) to one where it is much thinner (a five-seat federal commission currently operating with one member). Resolving the jurisdictional dispute doesn’t resolve the underlying governance question — it relocates it.
3 “The Gambler,” Kenny Rogers. Know when to hold ’em. Know when to fold ’em. Know when to walk away. The prediction market, in theory, will tell you. Whether to trust it is left as an exercise for the reader.
1 thought on “An Agenda-Setter With a Ticker”
Comments are closed.