A futures trader who checks an index quote before the cash open is treating a price as a forecast. Nobody surveyed anyone to produce that number. It came out of people committing money against each other, and it will move the instant one of them learns something the rest have not.

The same mechanism has been pointed at questions with nothing to do with rates or crude. Contracts that pay out on an election result, a court ruling or a Sunday football score have traded since 1988, and economists have spent almost four decades measuring whether the price on them says anything true about what people believe.

The short answer is yes, under conditions. The longer answer is the useful part, because the conditions and the failure modes were both on the record long before the current generation of online odds boards existed.

Hayek’s Point, Narrowed To One Dated Question

Friedrich Hayek’s 1945 essay in the American Economic Review, “The Use of Knowledge in Society,” argued that the knowledge an economy runs on never sits in one place. It is scattered across people who hold small, local pieces of it. The price system, in his account, is what moves those pieces into a single public number without anyone having to explain what they know.

An event contract shrinks that idea to a single sentence with a date and a settlement rule attached. Hold the contract, and it pays a fixed sum if the sentence turns out true and nothing if it does not. Because the price has to sit between those two payouts, it reads as a probability, and every trade adjusts it.

Justin Wolfers and Eric Zitzewitz wrote up that structure for the Journal of Economic Perspectives in 2004, surveying how far simple markets can turn dispersed information into a forecast.

A Poll Counts Heads, A Price Weighs Conviction

A poll draws a sample, asks a question, then weights the answers until the sample resembles the population. Its error is mostly a sampling problem, and a respondent pays nothing for a careless or dishonest answer.

A market skips representativeness altogether. Whoever turns up trades, so the pool is skewed by construction. What replaces the weighting scheme is conviction: a participant who thinks the quote is wrong can take a larger position, and someone who is only guessing has no reason to.

The costless-answer problem on the polling side keeps getting worse. Pew Research Center reported that its telephone survey response rates fell to 7% in 2017 and 6% in 2018, after several years near 9%.

Francis Galton showed the older version of the contrast in 1907, with 787 usable guesses at the dressed weight of an ox at a livestock fair. The median guess was 1,207 pounds and the animal came in at 1,198. Averaging works, but an average counts every ticket once. A market lets the person who actually knows something buy more of the answer.

Iowa Has Been Running The Experiment Since 1988

The Iowa Electronic Markets opened in 1988 and still trade. Faculty at the University of Iowa’s Henry B. Tippie College of Business run them as a teaching and research project, accounts open with $5 to $500 of real money, and the market operates under two staff no-action letters from the Commodity Futures Trading Commission dated February 1992 and June 1993.

Robert Forsythe, Forrest Nelson, George Neumann and Jack Wright published the first full anatomy of the 1988 market in the American Economic Review in 1992. Their result came in two halves. The market beat the opinion polls at calling the presidential outcome, and the traders inside it showed substantial judgment bias toward their own candidate.

The two halves fit together at the margin. Bias, they argued, describes average behavior, while price is set by whoever is willing to trade at the edge, and enough unbiased traders were active there to carry the number.

Joyce Berg, Nelson and Thomas Rietz then stretched the test across five presidential cycles. Set against 964 national polls from 1988 through 2004, the Iowa vote-share price landed closer to the eventual result 74% of the time, and it beat the polls in every election once the forecast was made more than 100 days out.

That horizon detail carries more weight than the headline share. A poll taken months before an election records an intention that has not formed yet. A price that far out is already discounting the campaign still to come.

The Yardstick Has Its Own Error Band

Every accuracy comparison is only as good as its benchmark, and the polling one has been audited by its own profession. The American Association for Public Opinion Research convened a task force on the 2020 pre-election polls and published its report in 2021.

Among polls fielded in the final two weeks, the average error on the margin came to 4.5 points for national popular-vote polls and 5.1 points for state-level presidential polls, and the task force called the national figure the largest in 40 years.

The error also ran in one direction. Polls in that window sat 3.9 points too favorable to Joe Biden nationally and 4.3 points too favorable in statewide presidential polls, and for Senate and governor races combined the overstatement of the Democratic margin reached 6.0 points.

None of that makes a market quote right. It does mean a price a few points away from the polling average is not automatically the outlier in the room.

Real Money Sharpens The Incentive, Not The Estimate

A common assumption holds that accuracy scales with the money involved. The experimental record disagrees. Emile Servan-Schreiber, Wolfers, David Pennock and Brian Galebach ran a real-money exchange and a play-money exchange side by side on NFL games through the 2003 to 2004 season and found the play-money forecasts performed as well as the cash ones.

The Iowa evidence points the same way, since the most studied accurate market in the literature caps a trader at a few hundred dollars. What appears to do the work is scoring. Participants are held to an outcome, and a careless position costs them something they care about, whether cash or standing. Deep capital helps a market absorb a large order, not form the estimate.

A Quote On A Chart Is Not The Same Thing As A Contract You Can Hold

A probability is only as precise as the sentence it settles against. Two venues can list what looks like the same question and resolve it differently, because one names a specific data source and a cutoff time and the other leaves the wording loose. Add the fee schedule and the bid-ask spread and the honest reading of any quote becomes a band, not a point.

That is why the practical question sits apart from the theoretical one. A screenshot of a price is a different thing from a position someone can actually open, since the second comes with a rulebook, a counterparty and a settlement source, and the best prediction markets coverage on LineUps works on that second ground, the venues themselves rather than the theory behind them.

Anyone setting a market price against a poll needs that venue detail first, because a contract that settles on a certified count is answering a different question from one that settles on a news call.

Thin Books Fail First, And They Show It Early

Aggregation needs something to aggregate. A contract with a dozen participants and a spread several points wide is one person’s opinion wearing a decimal point, and no theory upgrades it. Depth is the precondition, which is why a headline contract on a presidential race behaves nothing like one on a committee vote.

Calibration is the other half. A contract quoted at 30% is supposed to lose roughly seven times in ten. A single settled outcome therefore says almost nothing about whether the price was any good; only a long run of quotes against a long run of results does.

Longshots Have Been Overpriced For A Century

Betting markets carry a bias that has survived every dataset thrown at it. Longshots trade at prices implying they win more often than they do, and heavy favorites as if they win less often. The pattern was first documented in parimutuel horse racing and has since turned up in fixed-odds books and exchange pricing.

Erik Snowberg and Wolfers tested the competing explanations in the Journal of Political Economy in 2010 using a large database of American race starts. Their evidence favored misperception over appetite for risk: bettors misjudge small probabilities rather than knowingly paying for the thrill of a long price.

For anyone reading a quote, the consequence is narrow but firm. Prices near the tails are the least reliable part of the distribution, and they lean in a known direction.

Buying The Number Meets The Other Side Of The Book

The obvious objection to reading a price as a belief is that someone can buy the number. People have tried. Paul Rhode and Koleman Strumpf assembled a century of documented attempts to push American political betting odds and ran their own trades in the Iowa market, and the price effects they recorded were small and reversed quickly as other traders took the other side.

Robin Hanson, Ryan Oprea and David Porter put the same question in a laboratory market in 2006 and found that paid manipulators failed to degrade accuracy, because the rest of the room was rewarded for trading against them. Both results carry the thin-book caveat: the defense is a function of depth, and in a shallow contract the other side may not be there.

A contested contract reprices within minutes of a headline, while the reasoning behind the move takes hours to surface, so the commentary around a price tends to age better than the price did. Video suits that slower explanation better than a captured quote does, whether it comes from an exchange, an academic or a publisher such as LineUps.

Reading The Quote Without Over-Reading It

The regulatory position on event contracts in the United States is unsettled and moved again this summer. The Commodity Futures Trading Commission oversees designated contract markets, while state gaming regulators have argued that contracts written on sporting results fall under state gambling law.

On August 28, 2026, the Ninth Circuit held that one exchange had not shown federal law was likely to override Nevada’s rules for its sports contracts, and other courts have reached different results. Treat any flat statement about what is permitted, in either direction, as a position rather than a finding.

What the literature supports is narrower than the enthusiasm around it and more useful. A price on a dated question aggregates dispersed belief, it has beaten contemporaneous polls over a long sample at long horizons, and it does so through the marginal participant rather than the crowd average.

It also fails in specified ways: thin books, tail pricing, and resolution wording that does not say what a reader assumed. Nothing here is advice on taking a position, and a probability is not a plan. If staking money on an outcome has stopped feeling like a considered decision, the National Council on Problem Gambling’s helpline answers on 1-800-MY-RESET.