Part two of “The price of tomorrow“, a series on prediction and futures markets.
Part one of this series distinguished what futures markets and prediction markets are built to do. Futures markets transfer risk. Prediction markets aggregate belief. That distinction was framed as epistemological: one instrument manages an uncertain future, the other claims to know it. A question was left open at the close of that piece, and it deserves to be asked directly now. If a market can be built to price almost anything, including a war, a pandemic, or a presidential election, should it be?
Two charges, not one
The ethical case against futures markets is old: speculators distort the price of commodities real people depend on, and a trader who never intends to take delivery can still move the price a refinery pays, which moves through to the pump. The accusation carries genuine force. But it is a critique about distortion, not complicity. Futures markets concentrate financial power over goods that are not, in the end, purely financial.
Prediction markets invite a sharper accusation, because the object of the position differs in kind from the futures case. A futures trader who shorts wheat is indifferent to the harvest; she holds a financial position on a quantity, not an opinion about whether the harvest’s failure would be good or bad. A trader who buys “yes” on a contract asking whether a named city will suffer a terrorist attack has placed a bet whose payout is contingent on an event most people would regard as a moral catastrophe, not merely an economic one. That is complicity, or at least its appearance: a payout structure in which someone’s gain is mechanically tied to someone else’s tragedy. The mechanism, aggregating dispersed belief into a single price, is identical in both markets. The object of the wager is not. That is why the two critiques are not interchangeable, and treating them as one objection to “speculation” obscures more than it reveals.
Risk, uncertainty, and the limits of pricing
Frank Knight’s 1921 distinction between risk and uncertainty, refined separately by John Maynard Keynes in the Treatise on Probability, remains the sharpest tool available for this question. Risk is quantifiable: a known set of outcomes, an estimable probability distribution, a calculable expected value, the basis on which insurance and actuarial pricing operate. Uncertainty, in Knight’s stricter usage, resists this treatment. No stable distribution exists. No frequency table can be built. Keynes pressed the point further: for many of the questions that matter most, “we simply do not know.”
Futures markets were built for risk. A harvest’s size varies year to year, but the variation is bounded and statistically tractable; two centuries of agricultural data make next year’s wheat yield closer to a known distribution than an open question. Prediction markets, in their most ambitious applications, claim to price something closer to Knightian uncertainty itself: the probability of a war, a coup, a pandemic’s trajectory. These have no stable historical base rate, which is precisely what Knight and Keynes argued resists quantification by its nature, not merely by the limits of available data. A prediction market does not resolve this difficulty. It assigns the difficulty a number and trades it, and that number cannot claim the calibration of an election market, where outcomes are binary, dated, and drawn from a long historical series. The ambition is the same. The evidentiary base is not.
The Pentagon’s ghost
The clearest illustration of this tension is not Polymarket. It is a market that never opened.
In 2003, the Defense Advanced Research Projects Agency proposed the Policy Analysis Market: a prediction market, open to a curated set of participants, that would have allowed trading on contracts tied to geopolitical events in the Middle East, including the likelihood of coups, terrorist attacks, and assassinations. Two U.S. senators, Ron Wyden and Byron Dorgan, denounced the proposal within a day of its disclosure, calling it a “terror futures market,” and DARPA cancelled the program before a single contract traded. The idea that the federal government would operate a market in which someone could profit from correctly predicting an assassination was treated, correctly, as self-evidently grotesque.
Polymarket has since built, in practice and without government sponsorship, a structurally similar instrument. Geopolitical contracts on its platform have priced the probability of military strikes, ceasefire collapses, and leadership changes. No comparable Senate press conference followed. The reasons for that difference are not established by any data this piece can point to, but a plausible one is structural rather than moral: Polymarket carries no government sponsor, so there is no state actor whose conduct a senator can be seen to be policing by objecting to it. The absence of a state sponsor changes who is available to be held accountable, not whether the underlying ethical question has been answered. The structural ethical question, whether a financial incentive to be right about catastrophe is itself corrosive, was never resolved in 2003. It was simply never permitted to be asked of an instrument that had not yet been built. Two decades later, the instrument exists, the question remains open, and almost no one outside the trading itself is asking it.
What the discomfort gets wrong
The instinctive recoil from a market in catastrophe is not irrational. But it is worth testing against an uncomfortable counter-question: is the absence of such a market actually the safer or more ethical condition?
A pandemic that arrives with no market having priced its probability is not thereby a pandemic the world met with more foresight. It is a pandemic priced nowhere, met with whatever institutional capacity happened to exist independently of any market signal. A well-functioning market pricing pandemic probability is arguably a public good: it creates a financial incentive for someone to gather and act on dispersed information about zoonotic spillover risk before a slower institution does. Suppressing the market does not suppress the risk. It only suppresses the signal. Opacity about risk, on this reasoning, is not moral cleanliness. It is simply opacity.
The counterargument deserves its full weight, not a caricature of it. A market that pays out on catastrophe creates, at minimum, the appearance of a perverse incentive, and appearances govern trust even when they do not govern outcomes. A well-capitalised participant with both a large position and the capacity to influence the underlying event faces a temptation a passive bystander does not, which is the precise structural concern raised by part one‘s account of concentrated positioning in Polymarket’s 2024 election markets. A trader who can move the underlying probability, not merely the price that reflects it, has crossed from forecasting into something else. The anonymity that defines decentralised markets is precisely what makes that line nearly impossible to police from the outside.
The honest position is not that the discomfort is misplaced. It is that the discomfort, left unexamined, proves too much. It would, taken to its logical end, also indict insurance, which prices mortality and catastrophe daily without controversy, and credit default swaps, which price the probability of sovereign default. The relevant ethical distinction is not whether an instrument prices something grim. It is whether the instrument’s structure creates an incentive to cause the grim outcome rather than merely forecast it, and whether the market’s participants have any capacity to do so. That is a structural question, answerable case by case, not a blanket objection to the category.
But the structural version is not the strongest form the objection takes, and it would be a convenience to treat it as though it were. The harder version, the one Wyden and Dorgan were closer to articulating in 2003, does not depend on capacity to influence the outcome at all. It holds that pricing certain categories of event, an assassination, a mass-casualty attack, a genocide, is a moral injury independent of whether any trader could move the probability by a single basis point. On this view, the wrongness is not in the risk of manipulation. It is in the act of reducing a human catastrophe to a tradeable proposition at all, which treats the event as a commodity to be priced rather than a tragedy to be prevented, and does so regardless of how thin, how liquid, or how well-policed the market is. A market that priced the probability of a specific named individual’s assassination, run with perfect transparency, zero concentration risk, and full regulatory oversight, would on this view be no less objectionable than an opaque one. The object being priced is the problem, not the plumbing.
This version does not yield to the structural framework that follows. It cannot be answered by improving capacity controls or informational legitimacy, because it does not concede that those are the relevant variables. The honest acknowledgment is that this piece does not resolve it. The structural case made here addresses the conditions under which a prediction market’s operation can be unethical, manipulation, concentrated positioning, informational extraction, dressed up as forecasting. It does not address the prior question of whether some categories of event should never be priced regardless of how cleanly the operation runs. That prior question is closer to a question about what a society is willing to treat as a commodity than a question about market design, and it deserves to be named as a separate, unresolved objection rather than absorbed into the structural one as though answering the easier question disposes of the harder one.
Three questions before the trade
What follows from this is not a verdict on prediction markets as a class, and it does not purport to settle the deontological objection just raised. It is a narrower framework, sharpened by Polymarket but not limited to it, for an investor who has already concluded that pricing a given category of future event is not, in itself, impermissible, and who now needs to decide whether and how to engage with a specific instrument that does the pricing.
The first question is one of scale and capacity. Does a position in this market, at the size being considered, carry any plausible capacity to influence the underlying outcome rather than merely forecast it? A retail-sized position in a deep, liquid market does not. A concentrated position in a thin one might, and the Théo episode from part one is the clearest illustration of how quickly that line blurs in practice.
The second question is one of informational asymmetry and intent. Is the position built on genuinely dispersed information the market has not yet aggregated, or on private knowledge of the kind that would constitute an unfair advantage in any other market context? The Hayekian case for prediction markets depends on broad, honest participation. A position built on insider access to the outcome itself is not participation in price discovery. It is extraction from it.
The third question is one of consequence, asked plainly rather than rhetorically. If this position pays out, what does that imply happened in the world, and is the investor comfortable being the kind of market participant whose financial interest aligned with that outcome? This is not a legal test, and it is the question most likely to revive the deontological objection rather than dissolve it: an investor who works through the first two questions and finds no structural fault may still answer the third honestly and decline the trade. That is not a failure of the framework. It is the framework doing exactly what it is built to do, which is force the question rather than supply a default answer to it.
Where this leaves the question
These three questions do not resolve into a single answer, and the deeper objection they were built alongside, that some categories of event are not legitimately priceable at all, is not one they were built to answer. Any investor who works through them honestly may conclude that a given position is no more ethically fraught than a sovereign bond priced on default risk, or may conclude that no amount of structural cleanliness makes a market in human catastrophe something she wants her capital inside of. Neither conclusion is a failure to follow the framework. The second is a recognition that the framework has a boundary. The ethical question pricing the future raises is not whether the practice is permissible. It is whether the investor engaging in it has actually asked, and has been honest about which version of the question she has answered.
The “Price of tomorrow” continues next month with part three: “Reading the future, practically”, on how a investor should use prediction markets as a signal rather than a wager, and what the gap between Polymarket and futures-implied pricing has historically revealed.
Disclaimer: This article is for informational purposes only and does not constitute investment advice. Investors should conduct their own due diligence or consult with a financial advisor before making any investment decisions.
Photo by Massimiliano Morosinotto on Unsplash.
