What Is a Prediction Market? From 16th-Century Italy to Polymarket

The one-sentence version

A prediction market is a place where people trade contracts on whether something will happen, and the price of that contract — somewhere between zero and a dollar — is the crowd's real-time estimate of the probability. A contract sitting at 67 cents isn't a slogan, it's a claim: the market thinks there's about a 67% chance the event occurs, backed by people willing to put money behind that number.

The mechanics, without the jargon

Every prediction market question is written to have a clear, checkable answer — "Will X happen by Y date?" — and every contract on that question settles at exactly $1.00 if the answer turns out to be yes, or $0.00 if it's no. Buy at 30 cents and you're right, you make 70 cents on the dollar; buy at 30 cents and you're wrong, you lose your 30 cents entirely. Because the price moves constantly as new information and new traders show up, the market becomes a running, second-by-second poll — except instead of answering a survey for free, everyone participating has actual money on the line, which tends to sharpen how carefully people think before they weigh in.

Where the idea actually comes from

This isn't a crypto-era invention. Betting markets on political outcomes existed in Italian city-states as far back as the 16th century, and election betting was a genuinely active, semi-organized market in the United States from the late 1800s through around 1940, before it faded from public life. The modern, academically rigorous version of the idea started in 1988 at the University of Iowa, when a group of finance professors — Robert Forsythe, George Neumann, and Forrest Nelson among them — launched what's now called the Iowa Electronic Markets, originally as the Iowa Political Stock Market. The experiment tested something simple: does the crowd get sharper when real money, not just opinions, is on the line.

It worked strikingly well from the start. With only around 200 traders, mostly university staff, and individual positions capped at $500 under a CFTC no-action letter, the 1988 Iowa market forecast that George H.W. Bush would take 53.2% of the popular vote — close to the actual result — and put Michael Dukakis at 45.2% against his real 45.4%. A later academic review of the Iowa markets found they beat professional polling organizations about 74% of the time. That track record is the foundation everything after it was built on.

What happened between Iowa and today

The Hollywood Stock Exchange launched in 1996 as a play-money market where users traded virtual shares in movies and celebrities rather than real cash — proof that the mechanism worked even without financial stakes, though real money sharpens the incentives considerably. In 2004, HedgeStreet became the first prediction market to win approval from the CFTC as a Designated Contract Market under the Commodity Futures Modernization Act, setting the regulatory template Kalshi would eventually use sixteen years later. PredictIt followed a different, lighter-touch path, operating for years under a CFTC no-action letter rather than a full exchange license, and became a go-to source for political traders even with tighter position caps than the newer regulated platforms.

It's not just for elections

Companies use the same mechanism internally. Google ran internal prediction markets for years, letting employees trade on questions like product launch dates and other company-specific metrics; a 2015 academic analysis by Bo Cowgill and Eric Zitzewitz found those internal markets produced more accurate forecasts than the company's own subject-matter experts on the same questions. That's the same underlying logic as a public platform, just scoped to people inside one organization instead of the entire internet.

How you check whether a market is actually any good

Researchers grade prediction markets using something called a Brier score, which measures how well the predicted probabilities track what actually happens over many events. A well-calibrated market is one where, if you look at every contract that traded around 70 cents, roughly 70% of those events actually happened. That's a much higher bar than just "did it call the outcome right" — it's asking whether the market's confidence level itself can be trusted, which is what makes prediction markets useful as data rather than just a novelty scoreboard.

The clearest real-world test of that came during the 2024 US presidential election. Polymarket's implied probability for the eventual winner climbed past 60% in the final days, while several major polling-based forecasting models were still showing the race as close to a toss-up. The final result lined up more closely with where the markets had priced it, and that gap between market pricing and traditional polling is a big part of why prediction markets picked up so much mainstream and academic attention going into 2025 and 2026.

Where the wisdom of crowds breaks down

The whole idea rests on what's commonly called the wisdom of crowds — the idea that a large enough group with genuinely diverse information and opinions can out-forecast any single expert, a concept popularized well beyond economics circles. But that only holds under real conditions: enough liquidity that prices can actually move efficiently, a broad and genuinely independent set of participants, and a clearly defined question with an unambiguous resolution. Break any of those and the same mechanism that makes prediction markets useful can make them misleading — a thinly traded market can be pushed around by one large trader, and questions with fuzzy resolution criteria have caused real disputes on every major platform. Insider trading is the sharper version of the same problem: when someone trading a market already knows the answer, the price stops reflecting the crowd's genuine uncertainty and starts reflecting one person's private knowledge instead, which is exactly the pattern regulators and the platforms themselves have been chasing down as these markets have grown.

Where things stand now

What used to be a university research tool with $500 position caps is now a global category processing billions of dollars a month across politics, sports, weather, crypto, and geopolitics, split mainly between CFTC-regulated dollar exchanges like Kalshi and crypto-native platforms like Polymarket that have been working to build their own regulated US footholds. The core idea hasn't changed since 1988 — get enough people to put real money behind their honest estimate of a probability, and the resulting price tends to be worth paying attention to. What's changed is the scale, the speed, and how much attention regulators, journalists, and now entire industries are paying to the number that comes out the other end.

Disclaimer: This post is for informational and educational purposes only and is not financial or legal advice. Prediction markets involve real financial risk, calibration and accuracy can vary widely between markets, and legal status differs by platform, state, and country.